samples/CSP_with_element_substitution.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CSP using the crystal structure generator based on element substitution\n",
    "\n",
    "The following example predicts the stable crystal structure of a compound with its chemical composition. As described in our paper, the workflow consists of four steps:\n",
    "\n",
    "1. Find all possible template structures of the given query composition from a crystal structure database (Materials Project was used in this study).\n",
    "2. Replace elements in template structures with those in the query composition to generate a virtual library.\n",
    "3. Learn a surrogate model of the DFT energies, which is used to screen the virtual crystal structures according to the surrogate energies.\n",
    "4. Perform the DFT-based structure relaxation on a set of selected candidates.\n",
    "\n",
    "All these codes have been developed with Python 3.7 and [XenonPy](https://github.com/yoshida-lab/XenonPy) and tested with Python 3.7, 3.8, and 3.9. A DFT calculation backend is needed to execute the single-point DFT calculations and the DFT-based structure relaxations. In this study, [Vienna Ab initio Simulation Package (ak. VASP)](https://www.vasp.at/) was used.\n",
    "\n",
    "Please download the necessary files to run this demonstration.  All files should be organized as follows.\n",
    "\n",
    "```\n",
    "├── CSP_with_element_substitution.ipynb\n",
    "├── cgcnn_formation_energy\n",
    "│   ├── checkpoints\n",
    "│   │   ├── mae.pth.s\n",
    "│   │   ├── mse.pth.s\n",
    "│   │   ├── pearsonr.pth.s\n",
    "│   │   └── r2.pth.s\n",
    "│   ├── data_indices.pkl.z\n",
    "│   ├── describe.pkl.z\n",
    "│   ├── final_state.pth.s\n",
    "│   ├── ft_training_plot.png\n",
    "│   ├── ft_training_steps.png\n",
    "│   ├── init_state.pth.s\n",
    "│   ├── model.pth.m\n",
    "│   ├── model_class.pkl.z\n",
    "│   ├── model_params.pkl.z\n",
    "│   ├── model_structure.pkl.z\n",
    "│   ├── results.csv\n",
    "│   ├── results.pd.xz\n",
    "│   ├── training_info.pd.xz\n",
    "│   └── val_idx.pkl.z\n",
    "├── data\n",
    "│   ├── mp_structures_info.pd.xz\n",
    "│   └── searching_targets_with_predictions.pd.xz\n",
    "└── generated\n",
    "    ├── opt_candidate_result.pd.xz\n",
    "    └── single_point_results.pd.xz\n",
    "```\n",
    "\n",
    "* cgcnn_formation_energy: https://github.com/yoshida-lab/XenonPy/releases/download/v0.6.7/cgcnn_formation_energy.tar.gz\n",
    "* data: https://github.com/yoshida-lab/XenonPy/releases/download/v0.6.7/data.tar.gz\n",
    "* generated: https://github.com/yoshida-lab/XenonPy/releases/download/v0.6.7/generated.tar.gz"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Common packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib\n",
    "import json\n",
    "import operator\n",
    "\n",
    "from pathlib import Path\n",
    "from functools import reduce\n",
    "from collections import defaultdict\n",
    "\n",
    "\n",
    "from pymatgen.core import Structure, Element, Composition\n",
    "from pathlib import Path\n",
    "from matplotlib import pyplot as plt\n",
    "from matplotlib.ticker import FormatStrFormatter, PercentFormatter\n",
    "from joblib import Parallel, delayed, load\n",
    "from tqdm.notebook import tqdm\n",
    "\n",
    "\n",
    "# user-friendly print\n",
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "### 1. Load data\n",
    "\n",
    "In this study, we compiled a list of 90 stable crystalline compounds as the benchmark set (see our paper for details). All structures in the Materials Project database were used as a template set to create the virtual crystal structures to be screened."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>composition</th>\n",
       "      <th>elements</th>\n",
       "      <th>n_elemets</th>\n",
       "      <th>n_atoms</th>\n",
       "      <th>pretty_formula</th>\n",
       "      <th>space_group</th>\n",
       "      <th>space_group_number</th>\n",
       "      <th>volume</th>\n",
       "      <th>structure</th>\n",
       "      <th>full_formula</th>\n",
       "      <th>formation_energy</th>\n",
       "      <th>dataset</th>\n",
       "      <th>pred_volume</th>\n",
       "      <th>pred_spg</th>\n",
       "      <th>pred_spg_top20</th>\n",
       "      <th>rank</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>mp-569304</th>\n",
       "      <td>{'C': 4.0}</td>\n",
       "      <td>[C]</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>C</td>\n",
       "      <td>R-3m</td>\n",
       "      <td>166</td>\n",
       "      <td>62.601036</td>\n",
       "      <td>[[ 3.48460258  2.02633463 33.51650787] C, [1.0...</td>\n",
       "      <td>C4</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>14.695088</td>\n",
       "      <td>166</td>\n",
       "      <td>[166, 64, 194, 15, 62, 8, 160, 225, 221, 74, 1...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-149</th>\n",
       "      <td>{'Si': 2.0}</td>\n",
       "      <td>[Si]</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>Si</td>\n",
       "      <td>Fd-3m</td>\n",
       "      <td>227</td>\n",
       "      <td>40.888293</td>\n",
       "      <td>[[1.11629943 0.7893429  1.93348733] Si, [0. 0....</td>\n",
       "      <td>Si2</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>33.270061</td>\n",
       "      <td>225</td>\n",
       "      <td>[225, 227, 189, 194, 200, 129, 216, 191, 11, 1...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-2534</th>\n",
       "      <td>{'Ga': 1.0, 'As': 1.0}</td>\n",
       "      <td>[Ga, As]</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>GaAs</td>\n",
       "      <td>F-43m</td>\n",
       "      <td>216</td>\n",
       "      <td>47.531857</td>\n",
       "      <td>[[0. 0. 0.] Ga, [3.52125296 2.48990185 6.09898...</td>\n",
       "      <td>Ga1As1</td>\n",
       "      <td>-0.347249</td>\n",
       "      <td>1</td>\n",
       "      <td>41.831081</td>\n",
       "      <td>216</td>\n",
       "      <td>[216, 225, 160, 187, 166, 14, 5, 194, 44, 156,...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      composition  elements  n_elemets  n_atoms  \\\n",
       "mp-569304              {'C': 4.0}       [C]          1        4   \n",
       "mp-149                {'Si': 2.0}      [Si]          1        2   \n",
       "mp-2534    {'Ga': 1.0, 'As': 1.0}  [Ga, As]          2        2   \n",
       "\n",
       "          pretty_formula space_group  space_group_number     volume  \\\n",
       "mp-569304              C        R-3m                 166  62.601036   \n",
       "mp-149                Si       Fd-3m                 227  40.888293   \n",
       "mp-2534             GaAs       F-43m                 216  47.531857   \n",
       "\n",
       "                                                   structure full_formula  \\\n",
       "mp-569304  [[ 3.48460258  2.02633463 33.51650787] C, [1.0...           C4   \n",
       "mp-149     [[1.11629943 0.7893429  1.93348733] Si, [0. 0....          Si2   \n",
       "mp-2534    [[0. 0. 0.] Ga, [3.52125296 2.48990185 6.09898...       Ga1As1   \n",
       "\n",
       "           formation_energy  dataset  pred_volume  pred_spg  \\\n",
       "mp-569304          0.000000        1    14.695088       166   \n",
       "mp-149             0.000000        1    33.270061       225   \n",
       "mp-2534           -0.347249        1    41.831081       216   \n",
       "\n",
       "                                              pred_spg_top20  rank  \n",
       "mp-569304  [166, 64, 194, 15, 62, 8, 160, 225, 221, 74, 1...     0  \n",
       "mp-149     [225, 227, 189, 194, 200, 129, 216, 191, 11, 1...     1  \n",
       "mp-2534    [216, 225, 160, 187, 166, 14, 5, 194, 44, 156,...     0  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>full_formula</th>\n",
       "      <th>composition</th>\n",
       "      <th>composition_ratio</th>\n",
       "      <th>total_atoms</th>\n",
       "      <th>elements</th>\n",
       "      <th>n_elements</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>space_group</th>\n",
       "      <th>wy_cfg</th>\n",
       "      <th>wy_reformat</th>\n",
       "      <th>wy_pattern</th>\n",
       "      <th>wy_pattern_loose</th>\n",
       "      <th>wy_unique</th>\n",
       "      <th>ss_cfg</th>\n",
       "      <th>ss_reformat</th>\n",
       "      <th>ss_pattern</th>\n",
       "      <th>ss_pattern_loose</th>\n",
       "      <th>ss_unique</th>\n",
       "      <th>volume_of_cell</th>\n",
       "      <th>structure</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>mp-10018</th>\n",
       "      <td>Ac1</td>\n",
       "      <td>{'Ac': 1.0}</td>\n",
       "      <td>(1.0,)</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(Ac,)</td>\n",
       "      <td>1</td>\n",
       "      <td>225</td>\n",
       "      <td>Fm-3m</td>\n",
       "      <td>{'Ac': {'a': 4}}</td>\n",
       "      <td>{'Ac': ('a',)}</td>\n",
       "      <td>((a,),)</td>\n",
       "      <td>(a,)</td>\n",
       "      <td>(a,)</td>\n",
       "      <td>{'Ac': {'m-3m': 4}}</td>\n",
       "      <td>{'Ac': ('m-3m',)}</td>\n",
       "      <td>((m-3m,),)</td>\n",
       "      <td>(m-3m,)</td>\n",
       "      <td>(m-3m,)</td>\n",
       "      <td>45.384600</td>\n",
       "      <td>[[0. 0. 0.] Ac]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1006278</th>\n",
       "      <td>Ac1Eu1Au2</td>\n",
       "      <td>{'Ac': 1.0, 'Eu': 1.0, 'Au': 2.0}</td>\n",
       "      <td>(1.0, 1.0, 2.0)</td>\n",
       "      <td>4.0</td>\n",
       "      <td>(Ac, Au, Eu)</td>\n",
       "      <td>3</td>\n",
       "      <td>225</td>\n",
       "      <td>Fm-3m</td>\n",
       "      <td>{'Ac': {'b': 4}, 'Eu': {'a': 4}, 'Au': {'c': 8}}</td>\n",
       "      <td>{'Ac': ('b',), 'Eu': ('a',), 'Au': ('c',)}</td>\n",
       "      <td>((a,), (b,), (c,))</td>\n",
       "      <td>(a, b, c)</td>\n",
       "      <td>(a, b, c)</td>\n",
       "      <td>{'Ac': {'m-3m': 4}, 'Eu': {'m-3m': 4}, 'Au': {...</td>\n",
       "      <td>{'Ac': ('m-3m',), 'Eu': ('m-3m',), 'Au': ('-43...</td>\n",
       "      <td>((-43m,), (m-3m,), (m-3m,))</td>\n",
       "      <td>(-43m, m-3m, m-3m)</td>\n",
       "      <td>(-43m, m-3m)</td>\n",
       "      <td>117.080578</td>\n",
       "      <td>[[3.882859 3.882859 3.882859] Ac, [0. 0. 0.] E...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1008601</th>\n",
       "      <td>Zr1Ag2</td>\n",
       "      <td>{'Zr': 1.0, 'Ag': 2.0}</td>\n",
       "      <td>(1.0, 2.0)</td>\n",
       "      <td>3.0</td>\n",
       "      <td>(Ag, Zr)</td>\n",
       "      <td>2</td>\n",
       "      <td>139</td>\n",
       "      <td>I4/mmm</td>\n",
       "      <td>{'Zr': {'a': 2}, 'Ag': {'e': 4}}</td>\n",
       "      <td>{'Zr': ('a',), 'Ag': ('e',)}</td>\n",
       "      <td>((a,), (e,))</td>\n",
       "      <td>(a, e)</td>\n",
       "      <td>(a, e)</td>\n",
       "      <td>{'Zr': {'4/mmm': 2}, 'Ag': {'4mm': 4}}</td>\n",
       "      <td>{'Zr': ('4/mmm',), 'Ag': ('4mm',)}</td>\n",
       "      <td>((4/mmm,), (4mm,))</td>\n",
       "      <td>(4/mmm, 4mm)</td>\n",
       "      <td>(4/mmm, 4mm)</td>\n",
       "      <td>58.128751</td>\n",
       "      <td>[[0. 0. 0.] Zr, [0.         0.         6.08155...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           full_formula                        composition composition_ratio  \\\n",
       "id                                                                             \n",
       "mp-10018            Ac1                        {'Ac': 1.0}            (1.0,)   \n",
       "mp-1006278    Ac1Eu1Au2  {'Ac': 1.0, 'Eu': 1.0, 'Au': 2.0}   (1.0, 1.0, 2.0)   \n",
       "mp-1008601       Zr1Ag2             {'Zr': 1.0, 'Ag': 2.0}        (1.0, 2.0)   \n",
       "\n",
       "            total_atoms      elements  n_elements  space_group_num  \\\n",
       "id                                                                   \n",
       "mp-10018            1.0         (Ac,)           1              225   \n",
       "mp-1006278          4.0  (Ac, Au, Eu)           3              225   \n",
       "mp-1008601          3.0      (Ag, Zr)           2              139   \n",
       "\n",
       "           space_group                                            wy_cfg  \\\n",
       "id                                                                         \n",
       "mp-10018         Fm-3m                                  {'Ac': {'a': 4}}   \n",
       "mp-1006278       Fm-3m  {'Ac': {'b': 4}, 'Eu': {'a': 4}, 'Au': {'c': 8}}   \n",
       "mp-1008601      I4/mmm                  {'Zr': {'a': 2}, 'Ag': {'e': 4}}   \n",
       "\n",
       "                                           wy_reformat          wy_pattern  \\\n",
       "id                                                                           \n",
       "mp-10018                                {'Ac': ('a',)}             ((a,),)   \n",
       "mp-1006278  {'Ac': ('b',), 'Eu': ('a',), 'Au': ('c',)}  ((a,), (b,), (c,))   \n",
       "mp-1008601                {'Zr': ('a',), 'Ag': ('e',)}        ((a,), (e,))   \n",
       "\n",
       "           wy_pattern_loose  wy_unique  \\\n",
       "id                                       \n",
       "mp-10018               (a,)       (a,)   \n",
       "mp-1006278        (a, b, c)  (a, b, c)   \n",
       "mp-1008601           (a, e)     (a, e)   \n",
       "\n",
       "                                                       ss_cfg  \\\n",
       "id                                                              \n",
       "mp-10018                                  {'Ac': {'m-3m': 4}}   \n",
       "mp-1006278  {'Ac': {'m-3m': 4}, 'Eu': {'m-3m': 4}, 'Au': {...   \n",
       "mp-1008601             {'Zr': {'4/mmm': 2}, 'Ag': {'4mm': 4}}   \n",
       "\n",
       "                                                  ss_reformat  \\\n",
       "id                                                              \n",
       "mp-10018                                    {'Ac': ('m-3m',)}   \n",
       "mp-1006278  {'Ac': ('m-3m',), 'Eu': ('m-3m',), 'Au': ('-43...   \n",
       "mp-1008601                 {'Zr': ('4/mmm',), 'Ag': ('4mm',)}   \n",
       "\n",
       "                             ss_pattern    ss_pattern_loose     ss_unique  \\\n",
       "id                                                                          \n",
       "mp-10018                     ((m-3m,),)             (m-3m,)       (m-3m,)   \n",
       "mp-1006278  ((-43m,), (m-3m,), (m-3m,))  (-43m, m-3m, m-3m)  (-43m, m-3m)   \n",
       "mp-1008601           ((4/mmm,), (4mm,))        (4/mmm, 4mm)  (4/mmm, 4mm)   \n",
       "\n",
       "            volume_of_cell                                          structure  \n",
       "id                                                                             \n",
       "mp-10018         45.384600                                    [[0. 0. 0.] Ac]  \n",
       "mp-1006278      117.080578  [[3.882859 3.882859 3.882859] Ac, [0. 0. 0.] E...  \n",
       "mp-1008601       58.128751  [[0. 0. 0.] Zr, [0.         0.         6.08155...  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(126300, 20)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the benchmark set used in our paper\n",
    "searching_targets = pd.read_pickle('data/searching_targets_with_predictions.pd.xz')\n",
    "\n",
    "# template structures\n",
    "mp_structs_info = pd.read_pickle('data/mp_structures_info.pd.xz')\n",
    "mp_structs_info = mp_structs_info.assign(structure=mp_structs_info.structure.apply(lambda s: Structure.from_dict(s)))\n",
    "\n",
    "searching_targets.head(3)\n",
    "mp_structs_info.head(3)\n",
    "mp_structs_info.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For simplicity, we choose $\\mathrm{Zr}\\mathrm{O}_2$ as an example."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>composition</th>\n",
       "      <th>elements</th>\n",
       "      <th>n_elemets</th>\n",
       "      <th>n_atoms</th>\n",
       "      <th>pretty_formula</th>\n",
       "      <th>space_group</th>\n",
       "      <th>space_group_number</th>\n",
       "      <th>volume</th>\n",
       "      <th>structure</th>\n",
       "      <th>full_formula</th>\n",
       "      <th>formation_energy</th>\n",
       "      <th>dataset</th>\n",
       "      <th>pred_volume</th>\n",
       "      <th>pred_spg</th>\n",
       "      <th>pred_spg_top20</th>\n",
       "      <th>rank</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>mp-2858</th>\n",
       "      <td>{'Zr': 4.0, 'O': 8.0}</td>\n",
       "      <td>[Zr, O]</td>\n",
       "      <td>2</td>\n",
       "      <td>12</td>\n",
       "      <td>ZrO2</td>\n",
       "      <td>P2_1/c</td>\n",
       "      <td>14</td>\n",
       "      <td>147.110401</td>\n",
       "      <td>[[0.23276056 1.42220345 4.53138565] Zr, [2.401...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>-3.83322</td>\n",
       "      <td>1</td>\n",
       "      <td>140.843399</td>\n",
       "      <td>14</td>\n",
       "      <td>[14, 167, 87, 59, 189, 136, 141, 12, 92, 198, ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   composition elements  n_elemets  n_atoms pretty_formula  \\\n",
       "mp-2858  {'Zr': 4.0, 'O': 8.0}  [Zr, O]          2       12           ZrO2   \n",
       "\n",
       "        space_group  space_group_number      volume  \\\n",
       "mp-2858      P2_1/c                  14  147.110401   \n",
       "\n",
       "                                                 structure full_formula  \\\n",
       "mp-2858  [[0.23276056 1.42220345 4.53138565] Zr, [2.401...        Zr4O8   \n",
       "\n",
       "         formation_energy  dataset  pred_volume  pred_spg  \\\n",
       "mp-2858          -3.83322        1   140.843399        14   \n",
       "\n",
       "                                            pred_spg_top20  rank  \n",
       "mp-2858  [14, 167, 87, 59, 189, 136, 141, 12, 92, 198, ...     0  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "demo_target = 'ZrO2'\n",
    "\n",
    "demo_target = searching_targets[searching_targets.pretty_formula==demo_target]\n",
    "demo_target"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Drop the `demo_target` from the template pool."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(126299, 20)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>full_formula</th>\n",
       "      <th>composition</th>\n",
       "      <th>composition_ratio</th>\n",
       "      <th>total_atoms</th>\n",
       "      <th>elements</th>\n",
       "      <th>n_elements</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>space_group</th>\n",
       "      <th>wy_cfg</th>\n",
       "      <th>wy_reformat</th>\n",
       "      <th>wy_pattern</th>\n",
       "      <th>wy_pattern_loose</th>\n",
       "      <th>wy_unique</th>\n",
       "      <th>ss_cfg</th>\n",
       "      <th>ss_reformat</th>\n",
       "      <th>ss_pattern</th>\n",
       "      <th>ss_pattern_loose</th>\n",
       "      <th>ss_unique</th>\n",
       "      <th>volume_of_cell</th>\n",
       "      <th>structure</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>mp-10018</th>\n",
       "      <td>Ac1</td>\n",
       "      <td>{'Ac': 1.0}</td>\n",
       "      <td>(1.0,)</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(Ac,)</td>\n",
       "      <td>1</td>\n",
       "      <td>225</td>\n",
       "      <td>Fm-3m</td>\n",
       "      <td>{'Ac': {'a': 4}}</td>\n",
       "      <td>{'Ac': ('a',)}</td>\n",
       "      <td>((a,),)</td>\n",
       "      <td>(a,)</td>\n",
       "      <td>(a,)</td>\n",
       "      <td>{'Ac': {'m-3m': 4}}</td>\n",
       "      <td>{'Ac': ('m-3m',)}</td>\n",
       "      <td>((m-3m,),)</td>\n",
       "      <td>(m-3m,)</td>\n",
       "      <td>(m-3m,)</td>\n",
       "      <td>45.384600</td>\n",
       "      <td>[[0. 0. 0.] Ac]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1006278</th>\n",
       "      <td>Ac1Eu1Au2</td>\n",
       "      <td>{'Ac': 1.0, 'Eu': 1.0, 'Au': 2.0}</td>\n",
       "      <td>(1.0, 1.0, 2.0)</td>\n",
       "      <td>4.0</td>\n",
       "      <td>(Ac, Au, Eu)</td>\n",
       "      <td>3</td>\n",
       "      <td>225</td>\n",
       "      <td>Fm-3m</td>\n",
       "      <td>{'Ac': {'b': 4}, 'Eu': {'a': 4}, 'Au': {'c': 8}}</td>\n",
       "      <td>{'Ac': ('b',), 'Eu': ('a',), 'Au': ('c',)}</td>\n",
       "      <td>((a,), (b,), (c,))</td>\n",
       "      <td>(a, b, c)</td>\n",
       "      <td>(a, b, c)</td>\n",
       "      <td>{'Ac': {'m-3m': 4}, 'Eu': {'m-3m': 4}, 'Au': {...</td>\n",
       "      <td>{'Ac': ('m-3m',), 'Eu': ('m-3m',), 'Au': ('-43...</td>\n",
       "      <td>((-43m,), (m-3m,), (m-3m,))</td>\n",
       "      <td>(-43m, m-3m, m-3m)</td>\n",
       "      <td>(-43m, m-3m)</td>\n",
       "      <td>117.080578</td>\n",
       "      <td>[[3.882859 3.882859 3.882859] Ac, [0. 0. 0.] E...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1008601</th>\n",
       "      <td>Zr1Ag2</td>\n",
       "      <td>{'Zr': 1.0, 'Ag': 2.0}</td>\n",
       "      <td>(1.0, 2.0)</td>\n",
       "      <td>3.0</td>\n",
       "      <td>(Ag, Zr)</td>\n",
       "      <td>2</td>\n",
       "      <td>139</td>\n",
       "      <td>I4/mmm</td>\n",
       "      <td>{'Zr': {'a': 2}, 'Ag': {'e': 4}}</td>\n",
       "      <td>{'Zr': ('a',), 'Ag': ('e',)}</td>\n",
       "      <td>((a,), (e,))</td>\n",
       "      <td>(a, e)</td>\n",
       "      <td>(a, e)</td>\n",
       "      <td>{'Zr': {'4/mmm': 2}, 'Ag': {'4mm': 4}}</td>\n",
       "      <td>{'Zr': ('4/mmm',), 'Ag': ('4mm',)}</td>\n",
       "      <td>((4/mmm,), (4mm,))</td>\n",
       "      <td>(4/mmm, 4mm)</td>\n",
       "      <td>(4/mmm, 4mm)</td>\n",
       "      <td>58.128751</td>\n",
       "      <td>[[0. 0. 0.] Zr, [0.         0.         6.08155...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           full_formula                        composition composition_ratio  \\\n",
       "id                                                                             \n",
       "mp-10018            Ac1                        {'Ac': 1.0}            (1.0,)   \n",
       "mp-1006278    Ac1Eu1Au2  {'Ac': 1.0, 'Eu': 1.0, 'Au': 2.0}   (1.0, 1.0, 2.0)   \n",
       "mp-1008601       Zr1Ag2             {'Zr': 1.0, 'Ag': 2.0}        (1.0, 2.0)   \n",
       "\n",
       "            total_atoms      elements  n_elements  space_group_num  \\\n",
       "id                                                                   \n",
       "mp-10018            1.0         (Ac,)           1              225   \n",
       "mp-1006278          4.0  (Ac, Au, Eu)           3              225   \n",
       "mp-1008601          3.0      (Ag, Zr)           2              139   \n",
       "\n",
       "           space_group                                            wy_cfg  \\\n",
       "id                                                                         \n",
       "mp-10018         Fm-3m                                  {'Ac': {'a': 4}}   \n",
       "mp-1006278       Fm-3m  {'Ac': {'b': 4}, 'Eu': {'a': 4}, 'Au': {'c': 8}}   \n",
       "mp-1008601      I4/mmm                  {'Zr': {'a': 2}, 'Ag': {'e': 4}}   \n",
       "\n",
       "                                           wy_reformat          wy_pattern  \\\n",
       "id                                                                           \n",
       "mp-10018                                {'Ac': ('a',)}             ((a,),)   \n",
       "mp-1006278  {'Ac': ('b',), 'Eu': ('a',), 'Au': ('c',)}  ((a,), (b,), (c,))   \n",
       "mp-1008601                {'Zr': ('a',), 'Ag': ('e',)}        ((a,), (e,))   \n",
       "\n",
       "           wy_pattern_loose  wy_unique  \\\n",
       "id                                       \n",
       "mp-10018               (a,)       (a,)   \n",
       "mp-1006278        (a, b, c)  (a, b, c)   \n",
       "mp-1008601           (a, e)     (a, e)   \n",
       "\n",
       "                                                       ss_cfg  \\\n",
       "id                                                              \n",
       "mp-10018                                  {'Ac': {'m-3m': 4}}   \n",
       "mp-1006278  {'Ac': {'m-3m': 4}, 'Eu': {'m-3m': 4}, 'Au': {...   \n",
       "mp-1008601             {'Zr': {'4/mmm': 2}, 'Ag': {'4mm': 4}}   \n",
       "\n",
       "                                                  ss_reformat  \\\n",
       "id                                                              \n",
       "mp-10018                                    {'Ac': ('m-3m',)}   \n",
       "mp-1006278  {'Ac': ('m-3m',), 'Eu': ('m-3m',), 'Au': ('-43...   \n",
       "mp-1008601                 {'Zr': ('4/mmm',), 'Ag': ('4mm',)}   \n",
       "\n",
       "                             ss_pattern    ss_pattern_loose     ss_unique  \\\n",
       "id                                                                          \n",
       "mp-10018                     ((m-3m,),)             (m-3m,)       (m-3m,)   \n",
       "mp-1006278  ((-43m,), (m-3m,), (m-3m,))  (-43m, m-3m, m-3m)  (-43m, m-3m)   \n",
       "mp-1008601           ((4/mmm,), (4mm,))        (4/mmm, 4mm)  (4/mmm, 4mm)   \n",
       "\n",
       "            volume_of_cell                                          structure  \n",
       "id                                                                             \n",
       "mp-10018         45.384600                                    [[0. 0. 0.] Ac]  \n",
       "mp-1006278      117.080578  [[3.882859 3.882859 3.882859] Ac, [0. 0. 0.] E...  \n",
       "mp-1008601       58.128751  [[0. 0. 0.] Zr, [0.         0.         6.08155...  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "templates = mp_structs_info.loc[~mp_structs_info.index.isin(demo_target.index)]\n",
    "\n",
    "templates.shape\n",
    "templates.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Filter templates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# help functions to calculate dissimilarity and refine structures\n",
    "\n",
    "from matminer.featurizers.site import CrystalNNFingerprint\n",
    "from matminer.featurizers.structure import SiteStatsFingerprint\n",
    "from joblib import Parallel, delayed\n",
    "\n",
    "def calculate_similarity(anchor_struc, *struc, n_jobs=10):\n",
    "    # Calculate structure fingerprints.\n",
    "    ssf = SiteStatsFingerprint(\n",
    "        CrystalNNFingerprint.from_preset('ops', distance_cutoffs=None, x_diff_weight=0),\n",
    "        stats=('mean', 'std_dev', 'minimum', 'maximum'))\n",
    "    v_anchor = np.array(ssf.featurize(anchor_struc))\n",
    "    tmp = Parallel(n_jobs=n_jobs)(delayed(ssf.featurize)(s) for s in struc)\n",
    "    return [np.linalg.norm(np.array(s) - v_anchor) for s in tmp]\n",
    "\n",
    "def modify_struct(s: Structure, vol=None) -> Structure:\n",
    "    s = s.get_primitive_structure()\n",
    "    if vol is not None:\n",
    "        s.scale_lattice(vol)\n",
    "    return s"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For the given query composition $\\mathrm{Zr}\\mathrm{O}_2$, we collected a set of template crystal structures with the same composition ratio from the Materials Project database."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "demo_target = demo_target.iloc[0]\n",
    "\n",
    "# general info\n",
    "full_formula = demo_target.full_formula\n",
    "volume = demo_target.pred_volume\n",
    "spg_num = demo_target.space_group_number\n",
    "\n",
    "# composition info\n",
    "composition = demo_target.composition\n",
    "composition_ratio = tuple(sorted([v for v in composition.values()]))\n",
    "\n",
    "# groundtruth\n",
    "true_struct = demo_target.structure\n",
    "if spg_num not in [146, 148, 155, 160, 161, 166, 167]:\n",
    "    true_struct = SpacegroupAnalyzer(true_struct).get_conventional_standard_structure()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Filter templates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(924, 20)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>full_formula</th>\n",
       "      <th>composition</th>\n",
       "      <th>composition_ratio</th>\n",
       "      <th>total_atoms</th>\n",
       "      <th>elements</th>\n",
       "      <th>n_elements</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>space_group</th>\n",
       "      <th>wy_cfg</th>\n",
       "      <th>wy_reformat</th>\n",
       "      <th>wy_pattern</th>\n",
       "      <th>wy_pattern_loose</th>\n",
       "      <th>wy_unique</th>\n",
       "      <th>ss_cfg</th>\n",
       "      <th>ss_reformat</th>\n",
       "      <th>ss_pattern</th>\n",
       "      <th>ss_pattern_loose</th>\n",
       "      <th>ss_unique</th>\n",
       "      <th>volume_of_cell</th>\n",
       "      <th>structure</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>mp-1093999</th>\n",
       "      <td>Zn8Ag4</td>\n",
       "      <td>{'Zn': 8.0, 'Ag': 4.0}</td>\n",
       "      <td>(4.0, 8.0)</td>\n",
       "      <td>12.0</td>\n",
       "      <td>(Ag, Zn)</td>\n",
       "      <td>2</td>\n",
       "      <td>194</td>\n",
       "      <td>P6_3/mmc</td>\n",
       "      <td>{'Zn': {'a': 2, 'h': 6}, 'Ag': {'f': 4}}</td>\n",
       "      <td>{'Zn': ('a', 'h'), 'Ag': ('f',)}</td>\n",
       "      <td>((a, h), (f,))</td>\n",
       "      <td>(a, f, h)</td>\n",
       "      <td>(a, f, h)</td>\n",
       "      <td>{'Zn': {'-3m.': 2, 'mm2': 6}, 'Ag': {'3m.': 4}}</td>\n",
       "      <td>{'Zn': ('mm2', '-3m.'), 'Ag': ('3m.',)}</td>\n",
       "      <td>((3m.,), (mm2, -3m.))</td>\n",
       "      <td>(-3m., 3m., mm2)</td>\n",
       "      <td>(-3m., 3m., mm2)</td>\n",
       "      <td>140.843399</td>\n",
       "      <td>[[0. 0. 0.] Zn, [0.         0.         3.78643...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1094121</th>\n",
       "      <td>Mg4Ag8</td>\n",
       "      <td>{'Mg': 4.0, 'Ag': 8.0}</td>\n",
       "      <td>(4.0, 8.0)</td>\n",
       "      <td>12.0</td>\n",
       "      <td>(Ag, Mg)</td>\n",
       "      <td>2</td>\n",
       "      <td>194</td>\n",
       "      <td>P6_3/mmc</td>\n",
       "      <td>{'Mg': {'f': 4}, 'Ag': {'a': 2, 'h': 6}}</td>\n",
       "      <td>{'Mg': ('f',), 'Ag': ('a', 'h')}</td>\n",
       "      <td>((a, h), (f,))</td>\n",
       "      <td>(a, f, h)</td>\n",
       "      <td>(a, f, h)</td>\n",
       "      <td>{'Mg': {'3m.': 4}, 'Ag': {'-3m.': 2, 'mm2': 6}}</td>\n",
       "      <td>{'Mg': ('3m.',), 'Ag': ('-3m.', 'mm2')}</td>\n",
       "      <td>((-3m., mm2), (3m.,))</td>\n",
       "      <td>(-3m., 3m., mm2)</td>\n",
       "      <td>(-3m., 3m., mm2)</td>\n",
       "      <td>140.843399</td>\n",
       "      <td>[[2.34438871 1.35353592 0.45181891] Mg, [ 2.34...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mp-1095654</th>\n",
       "      <td>Si4Ag8</td>\n",
       "      <td>{'Si': 4.0, 'Ag': 8.0}</td>\n",
       "      <td>(4.0, 8.0)</td>\n",
       "      <td>12.0</td>\n",
       "      <td>(Ag, Si)</td>\n",
       "      <td>2</td>\n",
       "      <td>63</td>\n",
       "      <td>Cmcm</td>\n",
       "      <td>{'Si': {'f': 8}, 'Ag': {'a': 4, 'c': 4, 'g': 8}}</td>\n",
       "      <td>{'Si': ('f',), 'Ag': ('a', 'c', 'g')}</td>\n",
       "      <td>((a, c, g), (f,))</td>\n",
       "      <td>(a, c, f, g)</td>\n",
       "      <td>(a, c, f, g)</td>\n",
       "      <td>{'Si': {'m..': 8}, 'Ag': {'2/m..': 4, 'm2m': 4...</td>\n",
       "      <td>{'Si': ('m..',), 'Ag': ('..m', 'm2m', '2/m..')}</td>\n",
       "      <td>((..m, 2/m.., m2m), (m..,))</td>\n",
       "      <td>(..m, 2/m.., m.., m2m)</td>\n",
       "      <td>(..m, 2/m.., m.., m2m)</td>\n",
       "      <td>140.843399</td>\n",
       "      <td>[[ 2.35634937 -1.37790624  6.65122564] Si, [2....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           full_formula             composition composition_ratio  \\\n",
       "id                                                                  \n",
       "mp-1093999       Zn8Ag4  {'Zn': 8.0, 'Ag': 4.0}        (4.0, 8.0)   \n",
       "mp-1094121       Mg4Ag8  {'Mg': 4.0, 'Ag': 8.0}        (4.0, 8.0)   \n",
       "mp-1095654       Si4Ag8  {'Si': 4.0, 'Ag': 8.0}        (4.0, 8.0)   \n",
       "\n",
       "            total_atoms  elements  n_elements  space_group_num space_group  \\\n",
       "id                                                                           \n",
       "mp-1093999         12.0  (Ag, Zn)           2              194    P6_3/mmc   \n",
       "mp-1094121         12.0  (Ag, Mg)           2              194    P6_3/mmc   \n",
       "mp-1095654         12.0  (Ag, Si)           2               63        Cmcm   \n",
       "\n",
       "                                                      wy_cfg  \\\n",
       "id                                                             \n",
       "mp-1093999          {'Zn': {'a': 2, 'h': 6}, 'Ag': {'f': 4}}   \n",
       "mp-1094121          {'Mg': {'f': 4}, 'Ag': {'a': 2, 'h': 6}}   \n",
       "mp-1095654  {'Si': {'f': 8}, 'Ag': {'a': 4, 'c': 4, 'g': 8}}   \n",
       "\n",
       "                                      wy_reformat         wy_pattern  \\\n",
       "id                                                                     \n",
       "mp-1093999       {'Zn': ('a', 'h'), 'Ag': ('f',)}     ((a, h), (f,))   \n",
       "mp-1094121       {'Mg': ('f',), 'Ag': ('a', 'h')}     ((a, h), (f,))   \n",
       "mp-1095654  {'Si': ('f',), 'Ag': ('a', 'c', 'g')}  ((a, c, g), (f,))   \n",
       "\n",
       "           wy_pattern_loose     wy_unique  \\\n",
       "id                                          \n",
       "mp-1093999        (a, f, h)     (a, f, h)   \n",
       "mp-1094121        (a, f, h)     (a, f, h)   \n",
       "mp-1095654     (a, c, f, g)  (a, c, f, g)   \n",
       "\n",
       "                                                       ss_cfg  \\\n",
       "id                                                              \n",
       "mp-1093999    {'Zn': {'-3m.': 2, 'mm2': 6}, 'Ag': {'3m.': 4}}   \n",
       "mp-1094121    {'Mg': {'3m.': 4}, 'Ag': {'-3m.': 2, 'mm2': 6}}   \n",
       "mp-1095654  {'Si': {'m..': 8}, 'Ag': {'2/m..': 4, 'm2m': 4...   \n",
       "\n",
       "                                                ss_reformat  \\\n",
       "id                                                            \n",
       "mp-1093999          {'Zn': ('mm2', '-3m.'), 'Ag': ('3m.',)}   \n",
       "mp-1094121          {'Mg': ('3m.',), 'Ag': ('-3m.', 'mm2')}   \n",
       "mp-1095654  {'Si': ('m..',), 'Ag': ('..m', 'm2m', '2/m..')}   \n",
       "\n",
       "                             ss_pattern        ss_pattern_loose  \\\n",
       "id                                                                \n",
       "mp-1093999        ((3m.,), (mm2, -3m.))        (-3m., 3m., mm2)   \n",
       "mp-1094121        ((-3m., mm2), (3m.,))        (-3m., 3m., mm2)   \n",
       "mp-1095654  ((..m, 2/m.., m2m), (m..,))  (..m, 2/m.., m.., m2m)   \n",
       "\n",
       "                         ss_unique  volume_of_cell  \\\n",
       "id                                                   \n",
       "mp-1093999        (-3m., 3m., mm2)      140.843399   \n",
       "mp-1094121        (-3m., 3m., mm2)      140.843399   \n",
       "mp-1095654  (..m, 2/m.., m.., m2m)      140.843399   \n",
       "\n",
       "                                                    structure  \n",
       "id                                                             \n",
       "mp-1093999  [[0. 0. 0.] Zn, [0.         0.         3.78643...  \n",
       "mp-1094121  [[2.34438871 1.35353592 0.45181891] Mg, [ 2.34...  \n",
       "mp-1095654  [[ 2.35634937 -1.37790624  6.65122564] Si, [2....  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "filtered_templates = mp_structs_info[(mp_structs_info.composition_ratio == composition_ratio) & (mp_structs_info.composition != composition)]\n",
    "filtered_templates = filtered_templates.assign(\n",
    "    structure=filtered_templates.structure.apply(lambda s: modify_struct(s, vol=volume)),\n",
    "    volume_of_cell=[volume] * filtered_templates.shape[0],\n",
    ")\n",
    "\n",
    "filtered_templates.shape\n",
    "filtered_templates.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Perform postprocess to add dissimilarity and other information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Frequency'>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "filtered_templates = filtered_templates.assign(\n",
    "    dissimilarity=calculate_similarity(true_struct, *filtered_templates.structure),\n",
    "    target_formula=[full_formula] * filtered_templates.shape[0],\n",
    "    target_space_group_num=[spg_num] * filtered_templates.shape[0]\n",
    ").sort_values('dissimilarity')\n",
    "\n",
    "filtered_templates.dissimilarity.plot.hist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Considering that multiple crystals in the database belong to the same prototype structure (for example, 8,005 compounds have the same composition ratio A$_1$B$_1$C$_2$), a cluster-based template selection procedure was introduced. The objective is to select highly relevant templates with the query composition $\\mathrm{Zr}\\mathrm{O}_2$ while maintaining the diversity of the template structures. We applied DBSCAN to classify the templates into clusters in which the chemical compositions were converted into 290-dimensional compositional descriptors using XenonPy. Only the templates belonging to the same cluster as $\\mathrm{Zr}\\mathrm{O}_2$ were selected."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_615149/482518645.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  comps['target'] = demo_target.composition\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Target class number: 0\n",
      "  Estimated number of clusters: 12\n",
      "  Estimated number of noise points: 384\n"
     ]
    }
   ],
   "source": [
    "from xenonpy.descriptor import Compositions\n",
    "from sklearn.cluster import DBSCAN\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "desc_cal = Compositions()\n",
    "\n",
    "# Compute DBSCAN\n",
    "comps = filtered_templates.composition\n",
    "comps['target'] = demo_target.composition\n",
    "desc = desc_cal.fit_transform(comps)\n",
    "X = StandardScaler().fit_transform(desc)\n",
    "\n",
    "db = DBSCAN(eps=9, min_samples=10).fit(X)\n",
    "labels = db.labels_\n",
    "\n",
    "# Number of clusters in labels, ignoring noise if present.\n",
    "n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)\n",
    "n_noise_ = list(labels).count(-1)\n",
    "\n",
    "print('  Target class number: %d' % labels[-1])\n",
    "print('  Estimated number of clusters: %d' % n_clusters_)\n",
    "print('  Estimated number of noise points: %d' % n_noise_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(89, 23)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Frequency'>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "target_label = labels[-1]\n",
    "template_labels = labels[:-1]\n",
    "\n",
    "structs_ = filtered_templates.iloc[template_labels == target_label]\n",
    "structs_.shape\n",
    "structs_.dissimilarity.plot.hist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In addition, to eliminate structurally redundant templates, we used Pymatgen's StructureMatcher module to construct a unique set of templates that did not contain identical prototype structures."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "for spg: 1, size: 3\n",
      "for spg: 2, size: 7\n",
      "for spg: 4, size: 1\n",
      "for spg: 10, size: 2\n",
      "for spg: 11, size: 1\n",
      "for spg: 12, size: 9\n",
      "for spg: 14, size: 3\n",
      "for spg: 15, size: 2\n",
      "for spg: 18, size: 1\n",
      "for spg: 29, size: 1\n",
      "for spg: 44, size: 1\n",
      "for spg: 60, size: 4\n",
      "for spg: 62, size: 10\n",
      "for spg: 63, size: 8\n",
      "for spg: 70, size: 2\n",
      "for spg: 74, size: 6\n",
      "for spg: 87, size: 6\n",
      "for spg: 88, size: 1\n",
      "for spg: 111, size: 1\n",
      "for spg: 141, size: 1\n",
      "for spg: 160, size: 7\n",
      "for spg: 205, size: 6\n",
      "for spg: 227, size: 6\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(44, 4)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>structure</th>\n",
       "      <th>composition</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>dissimilarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[[2.82319719 4.36285554 3.72176206] Hf, [0.222...</td>\n",
       "      <td>{'Hf': 4.0, 'O': 8.0}</td>\n",
       "      <td>14</td>\n",
       "      <td>0.130907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[[1.3678768  1.24913405 0.18071245] Hf, [1.367...</td>\n",
       "      <td>{'Hf': 4.0, 'O': 8.0}</td>\n",
       "      <td>29</td>\n",
       "      <td>0.494103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[[3.67241649 1.45548662 1.50319392] Hf, [1.371...</td>\n",
       "      <td>{'Hf': 4.0, 'O': 8.0}</td>\n",
       "      <td>18</td>\n",
       "      <td>0.636794</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           structure            composition  \\\n",
       "0  [[2.82319719 4.36285554 3.72176206] Hf, [0.222...  {'Hf': 4.0, 'O': 8.0}   \n",
       "1  [[1.3678768  1.24913405 0.18071245] Hf, [1.367...  {'Hf': 4.0, 'O': 8.0}   \n",
       "2  [[3.67241649 1.45548662 1.50319392] Hf, [1.371...  {'Hf': 4.0, 'O': 8.0}   \n",
       "\n",
       "   space_group_num  dissimilarity  \n",
       "0               14       0.130907  \n",
       "1               29       0.494103  \n",
       "2               18       0.636794  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<Axes: ylabel='Frequency'>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from pymatgen.analysis.structure_matcher import StructureMatcher\n",
    "\n",
    "matcher = StructureMatcher(ltol=0.05, angle_tol=3)\n",
    "\n",
    "ret = []\n",
    "for spg_num_, ss in structs_.groupby('space_group_num'):\n",
    "    print(f'for spg: {spg_num_}, size: {ss.shape[0]}')\n",
    "    cand_structs = ss.structure\n",
    "\n",
    "    # final condidate\n",
    "    cand_structs = matcher.group_structures(cand_structs, anonymous=True)\n",
    "    for s in cand_structs:\n",
    "        ret.append([s[0]])\n",
    "\n",
    "if len(ret) == 0:\n",
    "    print('No templates')\n",
    "ret = pd.DataFrame(ret, columns=['structure'])\n",
    "filtered_templates = ret.assign(\n",
    "    composition=ret.structure.apply(lambda s: dict(s.composition.as_dict())),\n",
    "    space_group_num=ret.structure.apply(lambda s: s.get_space_group_info()[1]),\n",
    "    dissimilarity=calculate_similarity(true_struct, *ret.structure),\n",
    ").sort_values('dissimilarity').reset_index(drop=True)\n",
    "\n",
    "filtered_templates.shape\n",
    "filtered_templates.head(3)\n",
    "filtered_templates.dissimilarity.plot.hist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save templates for reusing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "filtered_templates.to_pickle('generated/filtered_templates.pd.xz')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Element substitution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>atomic_radius</th>\n",
       "      <th>atomic_radius_rahm</th>\n",
       "      <th>atomic_volume</th>\n",
       "      <th>atomic_weight</th>\n",
       "      <th>boiling_point</th>\n",
       "      <th>bulk_modulus</th>\n",
       "      <th>c6_gb</th>\n",
       "      <th>covalent_radius_cordero</th>\n",
       "      <th>covalent_radius_pyykko</th>\n",
       "      <th>covalent_radius_pyykko_double</th>\n",
       "      <th>...</th>\n",
       "      <th>num_s_unfilled</th>\n",
       "      <th>num_s_valence</th>\n",
       "      <th>specific_heat</th>\n",
       "      <th>thermal_conductivity</th>\n",
       "      <th>vdw_radius</th>\n",
       "      <th>vdw_radius_alvarez</th>\n",
       "      <th>vdw_radius_mm3</th>\n",
       "      <th>vdw_radius_uff</th>\n",
       "      <th>sound_velocity</th>\n",
       "      <th>Polarizability</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>H</th>\n",
       "      <td>79.000000</td>\n",
       "      <td>154.0</td>\n",
       "      <td>14.1</td>\n",
       "      <td>1.008000</td>\n",
       "      <td>20.280</td>\n",
       "      <td>56.79964</td>\n",
       "      <td>6.51</td>\n",
       "      <td>31.0</td>\n",
       "      <td>32.0</td>\n",
       "      <td>111.419933</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.122728</td>\n",
       "      <td>0.1805</td>\n",
       "      <td>110.0</td>\n",
       "      <td>120.0</td>\n",
       "      <td>162.0</td>\n",
       "      <td>288.6</td>\n",
       "      <td>1270.0</td>\n",
       "      <td>0.666793</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>He</th>\n",
       "      <td>147.832643</td>\n",
       "      <td>134.0</td>\n",
       "      <td>31.8</td>\n",
       "      <td>4.002602</td>\n",
       "      <td>4.216</td>\n",
       "      <td>85.10663</td>\n",
       "      <td>1.47</td>\n",
       "      <td>28.0</td>\n",
       "      <td>46.0</td>\n",
       "      <td>114.708355</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>5.188000</td>\n",
       "      <td>0.1513</td>\n",
       "      <td>140.0</td>\n",
       "      <td>143.0</td>\n",
       "      <td>153.0</td>\n",
       "      <td>236.2</td>\n",
       "      <td>970.0</td>\n",
       "      <td>0.205052</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Li</th>\n",
       "      <td>155.000000</td>\n",
       "      <td>220.0</td>\n",
       "      <td>13.1</td>\n",
       "      <td>6.940000</td>\n",
       "      <td>1118.150</td>\n",
       "      <td>11.00000</td>\n",
       "      <td>1410.00</td>\n",
       "      <td>128.0</td>\n",
       "      <td>133.0</td>\n",
       "      <td>124.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.489000</td>\n",
       "      <td>85.0000</td>\n",
       "      <td>182.0</td>\n",
       "      <td>212.0</td>\n",
       "      <td>255.0</td>\n",
       "      <td>245.1</td>\n",
       "      <td>6000.0</td>\n",
       "      <td>24.330000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 56 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    atomic_radius  atomic_radius_rahm  atomic_volume  atomic_weight  \\\n",
       "H       79.000000               154.0           14.1       1.008000   \n",
       "He     147.832643               134.0           31.8       4.002602   \n",
       "Li     155.000000               220.0           13.1       6.940000   \n",
       "\n",
       "    boiling_point  bulk_modulus    c6_gb  covalent_radius_cordero  \\\n",
       "H          20.280      56.79964     6.51                     31.0   \n",
       "He          4.216      85.10663     1.47                     28.0   \n",
       "Li       1118.150      11.00000  1410.00                    128.0   \n",
       "\n",
       "    covalent_radius_pyykko  covalent_radius_pyykko_double  ...  \\\n",
       "H                     32.0                     111.419933  ...   \n",
       "He                    46.0                     114.708355  ...   \n",
       "Li                   133.0                     124.000000  ...   \n",
       "\n",
       "    num_s_unfilled  num_s_valence  specific_heat  thermal_conductivity  \\\n",
       "H              1.0            1.0       1.122728                0.1805   \n",
       "He             0.0            2.0       5.188000                0.1513   \n",
       "Li             1.0            1.0       3.489000               85.0000   \n",
       "\n",
       "    vdw_radius  vdw_radius_alvarez  vdw_radius_mm3  vdw_radius_uff  \\\n",
       "H        110.0               120.0           162.0           288.6   \n",
       "He       140.0               143.0           153.0           236.2   \n",
       "Li       182.0               212.0           255.0           245.1   \n",
       "\n",
       "    sound_velocity  Polarizability  \n",
       "H           1270.0        0.666793  \n",
       "He           970.0        0.205052  \n",
       "Li          6000.0       24.330000  \n",
       "\n",
       "[3 rows x 56 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from xenonpy.datatools import preset\n",
    "\n",
    "elem_info = preset.elements_completed.drop(columns=['atomic_number', 'period'])\n",
    "\n",
    "elem_info.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "covalent_radius = {\n",
    "  \"H\": 0.26,\n",
    "  \"He\": 0.28,\n",
    "  \"Li\": 1.21,\n",
    "  \"Be\": 0.93,\n",
    "  \"B\": 0.81,\n",
    "  \"C\": 0.68,\n",
    "  \"N\": 0.7,\n",
    "  \"O\": 0.64,\n",
    "  \"F\": 0.54,\n",
    "  \"Ne\": 0.58,\n",
    "  \"Na\": 1.59,\n",
    "  \"Mg\": 1.34,\n",
    "  \"Al\": 1.18,\n",
    "  \"Si\": 1.09,\n",
    "  \"P\": 1.04,\n",
    "  \"S\": 1.02,\n",
    "  \"Cl\": 0.98,\n",
    "  \"Ar\": 0.96,\n",
    "  \"K\": 1.91,\n",
    "  \"Ca\": 1.86,\n",
    "  \"Sc\": 1.63,\n",
    "  \"Ti\": 1.52,\n",
    "  \"V\": 1.45,\n",
    "  \"Cr\": 1.34,\n",
    "  \"Mn\": 1.34,\n",
    "  \"Fe\": 1.29,\n",
    "  \"Co\": 1.23,\n",
    "  \"Ni\": 1.2,\n",
    "  \"Cu\": 1.28,\n",
    "  \"Zn\": 1.18,\n",
    "  \"Ga\": 1.19,\n",
    "  \"Ge\": 1.18,\n",
    "  \"As\": 1.15,\n",
    "  \"Se\": 1.16,\n",
    "  \"Br\": 1.17,\n",
    "  \"Kr\": 1.12,\n",
    "  \"Rb\": 2.11,\n",
    "  \"Sr\": 1.85,\n",
    "  \"Y\": 1.83,\n",
    "  \"Zr\": 1.68,\n",
    "  \"Nb\": 1.58,\n",
    "  \"Mo\": 1.49,\n",
    "  \"Tc\": 1.4,\n",
    "  \"Ru\": 1.39,\n",
    "  \"Rh\": 1.35,\n",
    "  \"Pd\": 1.33,\n",
    "  \"Ag\": 1.4,\n",
    "  \"Cd\": 1.35,\n",
    "  \"In\": 1.37,\n",
    "  \"Sn\": 1.35,\n",
    "  \"Sb\": 1.34,\n",
    "  \"Te\": 1.34,\n",
    "  \"I\": 1.36,\n",
    "  \"Xe\": 1.31,\n",
    "  \"Cs\": 2.33,\n",
    "  \"Ba\": 2.04,\n",
    "  \"La\": 1.99,\n",
    "  \"Ce\": 1.95,\n",
    "  \"Pr\": 1.96,\n",
    "  \"Nd\": 1.95,\n",
    "  \"Pm\": 1.99,\n",
    "  \"Sm\": 1.9,\n",
    "  \"Eu\": 1.93,\n",
    "  \"Gd\": 1.9,\n",
    "  \"Tb\": 1.89,\n",
    "  \"Dy\": 1.85,\n",
    "  \"Ho\": 1.85,\n",
    "  \"Er\": 1.83,\n",
    "  \"Tm\": 1.8,\n",
    "  \"Yb\": 1.79,\n",
    "  \"Lu\": 1.65,\n",
    "  \"Hf\": 1.81,\n",
    "  \"Ta\": 1.62,\n",
    "  \"W\": 1.55,\n",
    "  \"Re\": 1.44,\n",
    "  \"Os\": 1.4,\n",
    "  \"Ir\": 1.35,\n",
    "  \"Pt\": 1.31,\n",
    "  \"Au\": 1.33,\n",
    "  \"Hg\": 1.27,\n",
    "  \"Tl\": 1.38,\n",
    "  \"Pb\": 1.41,\n",
    "  \"Bi\": 1.44,\n",
    "  \"Po\": 1.36,\n",
    "  \"At\": 1.5,\n",
    "  \"Rn\": 1.5,\n",
    "  \"Fr\": 2.6,\n",
    "  \"Ra\": 2.19,\n",
    "  \"Ac\": 2.15,\n",
    "  \"Th\": 2.0,\n",
    "  \"Pa\": 2.0,\n",
    "  \"U\": 1.89,\n",
    "  \"Np\": 1.89,\n",
    "  \"Pu\": 1.86,\n",
    "  \"Am\": 1.74,\n",
    "  \"Cm\": 1.66\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from scipy.spatial.distance import cdist\n",
    "from itertools import product\n",
    "\n",
    "dis_matrix = cdist(elem_info, elem_info, 'seuclidean')\n",
    "dis_matrix = pd.DataFrame(dis_matrix, index=elem_info.index, columns=elem_info.index)\n",
    "\n",
    "def gen_replace_rule(tpl: dict, tar: dict) -> dict:\n",
    "    def extract(comp: dict) -> list:\n",
    "        ret = defaultdict(list)\n",
    "        for k, v in comp.items():\n",
    "            ret[v].append(k)\n",
    "\n",
    "        return [ret[k] for k in sorted(ret.keys())]\n",
    "    \n",
    "    tpl_pattern = extract(tpl)\n",
    "    tar_pattern = extract(tar)\n",
    "    \n",
    "    ret = {}\n",
    "    for tpl_e, tar_e in zip(tpl_pattern, tar_pattern):\n",
    "        if len(tpl) == 1:\n",
    "            ret[tpl_e[0]] = tar_e[0]\n",
    "        else:\n",
    "            dists = sorted([(dis_matrix.loc[e1, e2], e1, e2) for e1, e2 in product(tpl_e, tar_e)], key=lambda s: s[0])\n",
    "            while len(dists) > 0:\n",
    "                _, e1, e2 = dists.pop(0)\n",
    "                ret[e1] = e2\n",
    "                dists = [(d, e1_, e2_) for d, e1_, e2_ in dists if e1_ != e1 and e2_ != e2]\n",
    "    \n",
    "    return ret\n",
    "            \n",
    "\n",
    "def check_dis(struct, *, tot=0.1):\n",
    "    sites = struct.sites\n",
    "    dis_matrix = struct.distance_matrix\n",
    "    d = []\n",
    "\n",
    "    for s1, s2 in product(sites, sites):\n",
    "        d.append((covalent_radius[str(s1.specie)] + covalent_radius[str(s2.specie)]) * (1 - tot))\n",
    "    d = np.asarray(d)\n",
    "    d = d.reshape(int(np.sqrt(d.size)), -1)\n",
    "    np.fill_diagonal(d, 0.)\n",
    "    \n",
    "    return np.all((dis_matrix - d) >= 0)\n",
    "\n",
    "def gen(row, vol, g, target_formula):\n",
    "    s = row.structure.copy()\n",
    "    s.replace_species(rule)\n",
    "    s.scale_lattice(vol)\n",
    "    \n",
    "    if check_dis(s, tot=dis_tot):\n",
    "        return (s, target_formula, row.space_group_num, g, row.dissimilarity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "447924d8650940f9b31ad9f8eebe29ec",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Zr4O8:   0%|          | 0/44 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(5117, 5)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>structure</th>\n",
       "      <th>target_formula</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>structure_group</th>\n",
       "      <th>dissimilarity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3894</th>\n",
       "      <td>[[ 3.78213132e+00  2.67440362e+00 -3.47029717e...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>88</td>\n",
       "      <td>29</td>\n",
       "      <td>2.105175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4712</th>\n",
       "      <td>[[2.67819546 0.         0.        ] Zr, [0.   ...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>205</td>\n",
       "      <td>41</td>\n",
       "      <td>2.330398</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2206</th>\n",
       "      <td>[[5.40761366 1.19064409 3.29866046] Zr, [2.329...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>1.753520</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2965</th>\n",
       "      <td>[[2.73016113 2.73016113 0.        ] Zr, [2.730...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>205</td>\n",
       "      <td>8</td>\n",
       "      <td>1.907167</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4181</th>\n",
       "      <td>[[2.65952872 0.         0.        ] Zr, [0.   ...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>205</td>\n",
       "      <td>41</td>\n",
       "      <td>2.330398</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              structure target_formula  \\\n",
       "3894  [[ 3.78213132e+00  2.67440362e+00 -3.47029717e...          Zr4O8   \n",
       "4712  [[2.67819546 0.         0.        ] Zr, [0.   ...          Zr4O8   \n",
       "2206  [[5.40761366 1.19064409 3.29866046] Zr, [2.329...          Zr4O8   \n",
       "2965  [[2.73016113 2.73016113 0.        ] Zr, [2.730...          Zr4O8   \n",
       "4181  [[2.65952872 0.         0.        ] Zr, [0.   ...          Zr4O8   \n",
       "\n",
       "      space_group_num  structure_group  dissimilarity  \n",
       "3894               88               29       2.105175  \n",
       "4712              205               41       2.330398  \n",
       "2206                4                5       1.753520  \n",
       "2965              205                8       1.907167  \n",
       "4181              205               41       2.330398  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "n_structs = 1000\n",
    "volume_perturbation = 0.2\n",
    "dis_tot = 0.1\n",
    "n_jobs = 10\n",
    "\n",
    "pred_volume = demo_target.pred_volume\n",
    "target_comp = demo_target.composition\n",
    "target_formula = demo_target.full_formula\n",
    "\n",
    "ret = []\n",
    "for i, (idx, row) in tqdm(list(enumerate(filtered_templates.iterrows())), desc=full_formula):\n",
    "    rule = gen_replace_rule(row.composition, target_comp)\n",
    "    \n",
    "    ret += Parallel(n_jobs=n_jobs)(delayed(gen)(row, vol, i, target_formula) for vol in np.random.normal(pred_volume, pred_volume * volume_perturbation, size=n_structs))\n",
    "\n",
    "ret = [t for t in ret if t is not None]\n",
    "structure_candidate = pd.DataFrame(ret, columns=['structure', 'target_formula', 'space_group_num', 'structure_group', 'dissimilarity'])\n",
    "\n",
    "structure_candidate.shape\n",
    "structure_candidate.sample(5)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "Save generated virtual structures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "structure_candidate.to_pickle(f'generated/structure_candidate.pd.xz')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Calculate CGCNN descriptor and prepare instances for single point calculation\n",
    "\n",
    "Calculate CGCNN descriptors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from xenonpy.descriptor import CrystalGraphFeaturizer\n",
    "\n",
    "cg_featurizer = CrystalGraphFeaturizer(atom_feature='origin', radius=8, n_jobs=5)\n",
    "\n",
    "cg_features = cg_featurizer.transform(structure_candidate.structure)\n",
    "cg_features.to_pickle(str(f'generated/structure_candidate_cgcnn_features.pd'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select unopt structures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "N_UNOPT = 10\n",
    "\n",
    "single_point_candidate = structure_candidate.groupby('structure_group').sample(N_UNOPT, replace=True)\n",
    "single_point_candidate.to_pickle(f'generated/single_point_candidate.xz.pd')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here, users must perform single-point DFT energy calculations to create a training dataset of structure-energy instances in the transfer learning process. We assume that the DFT calculations have already been done and the results are saved in the file `single_point_results.xz.pd`.\n",
    "\n",
    "This sample code set provides a VASP calculation result from our study for those who are not comfortable performing DFT calculations. Please note that the structure data provided are not exactly the same as those generated above, as this code is only provided for demonstration purposes. Also, please note that only structures that have been successfully converged have been collected. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>structure</th>\n",
       "      <th>target_formula</th>\n",
       "      <th>space_group</th>\n",
       "      <th>space_group_num</th>\n",
       "      <th>dissimilarity</th>\n",
       "      <th>formation_energy_per_atom</th>\n",
       "      <th>final_energy_per_atom</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13345</th>\n",
       "      <td>[[0.84004836 1.40161604 5.96268943] Zr, [0.840...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pnma</td>\n",
       "      <td>62</td>\n",
       "      <td>0.992863</td>\n",
       "      <td>-3.327849</td>\n",
       "      <td>-9.475720</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13346</th>\n",
       "      <td>[[3.52539762 1.39721872 1.44301614] Zr, [1.316...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>P2_12_12</td>\n",
       "      <td>18</td>\n",
       "      <td>0.636794</td>\n",
       "      <td>-3.301080</td>\n",
       "      <td>-9.448950</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13347</th>\n",
       "      <td>[[0.92829469 1.17992972 1.91905367] Zr, [4.772...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>P1</td>\n",
       "      <td>1</td>\n",
       "      <td>2.121633</td>\n",
       "      <td>-0.954553</td>\n",
       "      <td>-7.102423</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13348</th>\n",
       "      <td>[[3.54745601 1.40596111 1.45204508] Zr, [1.324...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>P2_12_12</td>\n",
       "      <td>18</td>\n",
       "      <td>0.636794</td>\n",
       "      <td>-3.326914</td>\n",
       "      <td>-9.474785</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13349</th>\n",
       "      <td>[[0.82535534 1.37710081 5.85839787] Zr, [0.825...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pnma</td>\n",
       "      <td>62</td>\n",
       "      <td>0.992863</td>\n",
       "      <td>-3.305021</td>\n",
       "      <td>-9.452892</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13452</th>\n",
       "      <td>[[0.8479214  1.41475215 6.01857247] Zr, [0.847...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pnma</td>\n",
       "      <td>62</td>\n",
       "      <td>0.992863</td>\n",
       "      <td>-3.320972</td>\n",
       "      <td>-9.468843</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13453</th>\n",
       "      <td>[[1.32868312 1.2133427  0.17553451] Zr, [1.328...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pca2_1</td>\n",
       "      <td>29</td>\n",
       "      <td>0.494103</td>\n",
       "      <td>-3.373758</td>\n",
       "      <td>-9.521629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13454</th>\n",
       "      <td>[[1.29333833 1.18106613 0.17086506] Zr, [1.293...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pca2_1</td>\n",
       "      <td>29</td>\n",
       "      <td>0.494103</td>\n",
       "      <td>-3.199693</td>\n",
       "      <td>-9.347563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13455</th>\n",
       "      <td>[[0.         5.9796572  2.03446975] Zr, [2.051...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>C2/c</td>\n",
       "      <td>15</td>\n",
       "      <td>2.092321</td>\n",
       "      <td>-1.922489</td>\n",
       "      <td>-8.070360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13456</th>\n",
       "      <td>[[1.31199976 1.1981076  0.17333045] Zr, [1.311...</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>Pca2_1</td>\n",
       "      <td>29</td>\n",
       "      <td>0.494103</td>\n",
       "      <td>-3.309452</td>\n",
       "      <td>-9.457323</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>112 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               structure target_formula  \\\n",
       "13345  [[0.84004836 1.40161604 5.96268943] Zr, [0.840...          Zr4O8   \n",
       "13346  [[3.52539762 1.39721872 1.44301614] Zr, [1.316...          Zr4O8   \n",
       "13347  [[0.92829469 1.17992972 1.91905367] Zr, [4.772...          Zr4O8   \n",
       "13348  [[3.54745601 1.40596111 1.45204508] Zr, [1.324...          Zr4O8   \n",
       "13349  [[0.82535534 1.37710081 5.85839787] Zr, [0.825...          Zr4O8   \n",
       "...                                                  ...            ...   \n",
       "13452  [[0.8479214  1.41475215 6.01857247] Zr, [0.847...          Zr4O8   \n",
       "13453  [[1.32868312 1.2133427  0.17553451] Zr, [1.328...          Zr4O8   \n",
       "13454  [[1.29333833 1.18106613 0.17086506] Zr, [1.293...          Zr4O8   \n",
       "13455  [[0.         5.9796572  2.03446975] Zr, [2.051...          Zr4O8   \n",
       "13456  [[1.31199976 1.1981076  0.17333045] Zr, [1.311...          Zr4O8   \n",
       "\n",
       "      space_group space_group_num  dissimilarity  formation_energy_per_atom  \\\n",
       "13345        Pnma              62       0.992863                  -3.327849   \n",
       "13346    P2_12_12              18       0.636794                  -3.301080   \n",
       "13347          P1               1       2.121633                  -0.954553   \n",
       "13348    P2_12_12              18       0.636794                  -3.326914   \n",
       "13349        Pnma              62       0.992863                  -3.305021   \n",
       "...           ...             ...            ...                        ...   \n",
       "13452        Pnma              62       0.992863                  -3.320972   \n",
       "13453      Pca2_1              29       0.494103                  -3.373758   \n",
       "13454      Pca2_1              29       0.494103                  -3.199693   \n",
       "13455        C2/c              15       2.092321                  -1.922489   \n",
       "13456      Pca2_1              29       0.494103                  -3.309452   \n",
       "\n",
       "       final_energy_per_atom  \n",
       "13345              -9.475720  \n",
       "13346              -9.448950  \n",
       "13347              -7.102423  \n",
       "13348              -9.474785  \n",
       "13349              -9.452892  \n",
       "...                      ...  \n",
       "13452              -9.468843  \n",
       "13453              -9.521629  \n",
       "13454              -9.347563  \n",
       "13455              -8.070360  \n",
       "13456              -9.457323  \n",
       "\n",
       "[112 rows x 7 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "single_point_results = pd.read_pickle('generated/single_point_results.pd.xz')\n",
    "single_point_results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Train the surrogate model\n",
    "\n",
    "As described in our paper, a transferred energy prediction model was used to perform exhaustive virtual screening. The global energy prediction model, the CGCNN model was trained on 126,210 stable and unstable crystal structures, with their formation energies retrieved from the Materials Project; the 90 benchmark crystals were excluded from the training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from shutil import copyfile\n",
    "from pathlib import Path\n",
    "from itertools import product\n",
    "from collections import OrderedDict\n",
    "from datetime import datetime, timedelta\n",
    "from platform import version as sys_ver\n",
    "from sys import version as py_ver\n",
    "\n",
    "import torch\n",
    "import warnings\n",
    "from tqdm import tqdm\n",
    "\n",
    "import matplotlib.cm as mplcm\n",
    "import matplotlib.colors as colors\n",
    "\n",
    "from torch.utils.data import DataLoader\n",
    "from xenonpy.descriptor import Compositions, CrystalGraphFeaturizer\n",
    "from xenonpy.datatools import Splitter\n",
    "from xenonpy.utils.math import Product\n",
    "\n",
    "from xenonpy.model.utils import regression_metrics\n",
    "from xenonpy.model.training import Trainer, SGD, MSELoss, Adam, ReduceLROnPlateau, ExponentialLR, ClipNorm, Checker, ClipValue\n",
    "from xenonpy.model.training.extension import Validator, TensorConverter, Persist\n",
    "from xenonpy.model.training.dataset import ArrayDataset, CrystalGraphDataset\n",
    "from xenonpy.model import CrystalGraphConvNet\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from xenonpy.model.utils.metrics import regression_metrics\n",
    "\n",
    "def cv_plot(pred, true, pred_fit=None, true_fit=None, *, unit='', lim=None, title='', ax=None, style='seaborn-notebook', **kwargs):\n",
    "    pred, true = pred.flatten(), true.flatten()\n",
    "    scores = regression_metrics(true, pred)\n",
    "\n",
    "    with matplotlib.style.context(style):\n",
    "        if ax is None:\n",
    "            _, ax = plt.subplots(figsize=(8, 8), dpi=100)\n",
    "\n",
    "        if pred_fit is not None and true_fit is not None:\n",
    "            ax.scatter(pred_fit, true_fit, alpha=0.5, s=5, label='Train', **kwargs)\n",
    "        ax.scatter(pred, true, alpha=0.6, s=8, label='Test', **kwargs)\n",
    "        if lim is not None:\n",
    "            ax.set_xlim(*lim)\n",
    "            ax.set_ylim(*lim)\n",
    "        else:\n",
    "            lim = ax.get_xlim()\n",
    "            ax.set_xlim(*lim)\n",
    "            ax.set_ylim(*lim)\n",
    "\n",
    "        ax.plot(lim, lim, ls=\"--\", c=\"0.3\", alpha=0.7, lw=1)\n",
    "        if unit != '':\n",
    "            unit = f' ({unit})'\n",
    "        ax.set_xlabel(f'Prediction{unit}', fontsize='x-large')\n",
    "        ax.set_ylabel(f'Observation{unit}', fontsize='x-large')\n",
    "        legend = ax.legend(markerscale=2, fontsize='larger', loc=0)\n",
    "        for lh in legend.legendHandles:\n",
    "            lh.set_alpha(1.0)\n",
    "        ax.grid(color='gray', linestyle='--', linewidth=0.5, alpha=0.5)\n",
    "        ax.set_title(title, fontsize='x-large')\n",
    "        shift = (lim[1] - lim[0]) / 50\n",
    "        ax.text(lim[1] - shift, lim[0] + shift,\n",
    "                 'R2: %.5f\\nMAE: %.5f\\nRMSE: %.5f\\nPearsonR: %.5f\\nSpearmanR: %.5f' % (scores['r2'],scores['mae'], scores['rmse'], scores['pearsonr'], scores['spearmanr']),\n",
    "                 horizontalalignment='right', verticalalignment='bottom', fontsize='larger',  bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.5))\n",
    "        ax.tick_params(axis='both', which='major', labelsize='larger')\n",
    "        #     plt.show()\n",
    "        return ax"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Prepare CGCNN descriptors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "cg_featurizer = CrystalGraphFeaturizer(atom_feature='origin', radius=8, n_jobs=1)\n",
    "\n",
    "\n",
    "#### calculate cgcnn feartures\n",
    "cg_features = cg_featurizer.transform(single_point_results.structure)\n",
    "\n",
    "## prepare training\n",
    "data_size = cg_features.shape[0]\n",
    "if data_size <= 10:\n",
    "    splitter = Splitter(data_size, test_size=2, random_state=100)\n",
    "else:\n",
    "    splitter = Splitter(data_size, test_size=0.1, random_state=100)\n",
    "x_train, x_test, y_train, y_test = splitter.split(cg_features, single_point_results.formation_energy_per_atom.to_frame())\n",
    "train_dataloader = DataLoader(CrystalGraphDataset(x_train, y_train), shuffle=False,\n",
    "                              batch_size=256, collate_fn=CrystalGraphDataset.collate_fn)\n",
    "val_dataloader = DataLoader(CrystalGraphDataset(x_test, y_test),\n",
    "                              batch_size=2000, collate_fn=CrystalGraphDataset.collate_fn)\n",
    "\n",
    "## set training parameters\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "##### Fine-tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# global model\n",
    "model_path = 'cgcnn_formation_energy'\n",
    "\n",
    "# fine-tuning parameter\n",
    "params = [0.01, 0.008, 0.006, 0.004, 0.002]\n",
    "\n",
    "# max epochs\n",
    "epochs = 400"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "====== lr: 0.006; clip: 0.006 ===========\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "49632002492848ca88c5fa6547aa7340",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training:   0%|          | 0/400 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Early stopping is applied: no improvement for ['pearsonr', 'mse', 'r2', 'mae'] since the last 21 iterations, finish training at iteration 179\n",
      "====== lr: 0.006; clip: 0.004 ===========\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4e1353376b4f47478d0230fd3304c05c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training:   0%|          | 0/400 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Early stopping is applied: no improvement for ['pearsonr', 'mse', 'r2', 'mae'] since the last 21 iterations, finish training at iteration 245\n",
      "====== lr: 0.004; clip: 0.006 ===========\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "060809897a7a4ae08c1a8c9ba1583fd9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training:   0%|          | 0/400 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Early stopping is applied: no improvement for ['pearsonr', 'mse', 'r2', 'mae'] since the last 21 iterations, finish training at iteration 174\n",
      "====== lr: 0.004; clip: 0.004 ===========\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "443c0b845c6e4d1e85f19fe2cf60d011",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training:   0%|          | 0/400 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Early stopping is applied: no improvement for ['pearsonr', 'mse', 'r2', 'mae'] since the last 21 iterations, finish training at iteration 146\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'generated/transferred_model/ft_training_best_mae.png'"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "summary = []\n",
    "\n",
    "## model training with grid search\n",
    "for lr, clip in product(params, params):\n",
    "\n",
    "    checker = Checker(model_path)\n",
    "    trainer = Trainer.from_checker(\n",
    "        checker=checker,\n",
    "        cuda=True,\n",
    "        optimizer=Adam(lr=lr),\n",
    "        loss_func=MSELoss(),\n",
    "        clip_grad=ClipValue(clip)\n",
    "    ).extend(\n",
    "        TensorConverter(empty_cache=True),\n",
    "        Validator(metrics_func=regression_metrics, early_stopping=20, trace_order=1, pearsonr=1.0, mse=0.0, r2=0.0, mae=0.0),\n",
    "    )\n",
    "\n",
    "    persist = Persist(\n",
    "        increment=True,\n",
    "        path=f'generated/transferred_model/model',\n",
    "        model_class=CrystalGraphConvNet,\n",
    "        model_params=checker.model_params,\n",
    "        author='Chang Liu',\n",
    "        email='liu.chang@ism.ac.jp',\n",
    "        dataset_detail='vasp-6.1.2'\n",
    "    )\n",
    "    trainer.extend(persist).reset(to='mae')\n",
    "\n",
    "\n",
    "    print(f\"====== lr: {lr}; clip: {clip} ===========\")\n",
    "    trainer.fit(training_dataset=train_dataloader, validation_dataset=val_dataloader, epochs=epochs, checkpoint=True)\n",
    "\n",
    "    persist(splitter=splitter, data_indices=single_point_results.index.tolist())\n",
    "    training_info = trainer.training_info\n",
    "\n",
    "    # draw training steps\n",
    "    f, ax = plt.subplots(figsize=(10, 7),dpi=100)\n",
    "    ax = trainer.training_info.plot(y=['train_mse_loss', 'val_mse'], ax=ax)\n",
    "    _ = ax.grid(axis='y', linestyle='--')\n",
    "    f.savefig(f'{persist.path}/ft_training_steps.png', bbox_inches='tight', dpi=300)\n",
    "\n",
    "    # draw OvP\n",
    "    pred, true = trainer.predict(dataset=val_dataloader, checkpoint='mae')\n",
    "    pred_fit, true_fit = trainer.predict(dataset=train_dataloader, checkpoint='mae')\n",
    "\n",
    "    _ = cv_plot(pred, true, pred_fit, true_fit)\n",
    "    plt.savefig(f'{persist.path}/ft_training_plot.png', bbox_inches='tight', dpi=300)\n",
    "    plt.cla()\n",
    "    plt.clf()\n",
    "    plt.close()\n",
    "\n",
    "    summary.append(dict(\n",
    "        path=persist.path,\n",
    "        data_size=data_size,\n",
    "        lr=lr,\n",
    "        clip=clip,\n",
    "        mae=training_info['val_mae'].min(),\n",
    "        mse=training_info['val_mse'].min(),\n",
    "        r2=training_info['val_r2'].max(),\n",
    "        corr=training_info['val_pearsonr'].max(),\n",
    "    ))\n",
    "    pd.DataFrame(summary).to_csv('generated/transferred_model/training_summary.pd.csv')\n",
    "\n",
    "summary = pd.DataFrame(summary)\n",
    "summary.to_pickle('generated/transferred_model/training_summary.pd.xz')\n",
    "\n",
    "best_mae = summary.sort_values('mae').iloc[0].path\n",
    "copyfile(f'{best_mae}/ft_training_plot.png', 'generated/transferred_model/ft_training_best_mae.png')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Prediction and virtual screening"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "N_GROUP = 20\n",
    "N_OPT = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(89, 6)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_mae_path = pd.read_pickle('generated/transferred_model/training_summary.pd.xz').sort_values('mae').iloc[0].path\n",
    "trainer = Trainer.from_checker(checker=best_mae_path, cuda=True).extend(TensorConverter(empty_cache=True))\n",
    "\n",
    "### prediction\n",
    "screen_cgcnn_features = pd.read_pickle('generated/structure_candidate_cgcnn_features.pd')\n",
    "val_dataloader = DataLoader(CrystalGraphDataset(screen_cgcnn_features, np.zeros(screen_cgcnn_features.shape[0]).reshape(-1, 1)),\n",
    "                              batch_size=500, collate_fn=CrystalGraphDataset.collate_fn)\n",
    "\n",
    "pred, _ = trainer.predict(dataset=val_dataloader, checkpoint='mae')\n",
    "\n",
    "### load screen structures\n",
    "screen_structures = pd.read_pickle('generated/structure_candidate.pd.xz')\n",
    "screen_structures = screen_structures.assign(prediction=pred)\n",
    "screen_structures.drop(columns='structure').to_pickle('generated/structure_candidate_with_prediction.pd.xz')\n",
    "\n",
    "screen_structures = screen_structures[screen_structures.prediction < 0]\n",
    "groups = screen_structures.groupby('structure_group')[['dissimilarity', 'prediction']].min().sort_values('prediction').index[:N_GROUP]\n",
    "opt_candidate = screen_structures[screen_structures.structure_group.isin(groups)].groupby('structure_group').apply(lambda x: x.sort_values('prediction').head(N_OPT)).droplevel('structure_group')\n",
    "opt_candidate.to_pickle('generated/opt_candidate.pd.xz')\n",
    "\n",
    "opt_candidate.shape\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Selection of the final candidates\n",
    "\n",
    "The DFT structural optimization is performed on the candidate structures saved in `opt_candidate.pd.xz` to predict the crystal structure of $\\mathrm{Zr}\\mathrm{O}_2$. The structure optimization should be performed by the user."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(37, 7)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>formula_pretty</th>\n",
       "      <th>formation_energy_per_atom</th>\n",
       "      <th>final_energy_per_atom</th>\n",
       "      <th>structure_group</th>\n",
       "      <th>target_formula</th>\n",
       "      <th>structure</th>\n",
       "      <th>converged</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1059</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901437</td>\n",
       "      <td>-9.438405</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61332090e+00 2.61332090e+00 3.20039507e-16...</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>967</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901465</td>\n",
       "      <td>-9.438433</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61330891e+00 2.61330891e+00 3.20038039e-16...</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>65</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982247</td>\n",
       "      <td>-9.519214</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.66157261 1.3074229  1.49870806] Zr, [1.370...</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     formula_pretty  formation_energy_per_atom  final_energy_per_atom  \\\n",
       "id                                                                      \n",
       "1059           ZrO2                  -7.901437              -9.438405   \n",
       "967            ZrO2                  -7.901465              -9.438433   \n",
       "65             ZrO2                  -7.982247              -9.519214   \n",
       "\n",
       "      structure_group target_formula  \\\n",
       "id                                     \n",
       "1059                8          Zr4O8   \n",
       "967                 8          Zr4O8   \n",
       "65                  2          Zr4O8   \n",
       "\n",
       "                                              structure  converged  \n",
       "id                                                                  \n",
       "1059  [[2.61332090e+00 2.61332090e+00 3.20039507e-16...       True  \n",
       "967   [[2.61330891e+00 2.61330891e+00 3.20038039e-16...       True  \n",
       "65    [[3.66157261 1.3074229  1.49870806] Zr, [1.370...       True  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt_candidate_result = pd.read_pickle('generated/opt_candidate_result.pd.xz')\n",
    "\n",
    "opt_candidate_result.shape\n",
    "opt_candidate_result.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The top 5 structures with the lowest formation energy for each template are selected as the final candidates."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>formula_pretty</th>\n",
       "      <th>formation_energy_per_atom</th>\n",
       "      <th>final_energy_per_atom</th>\n",
       "      <th>structure_group</th>\n",
       "      <th>target_formula</th>\n",
       "      <th>structure</th>\n",
       "      <th>converged</th>\n",
       "      <th>dissimilarity</th>\n",
       "      <th>space_group_num</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.056902</td>\n",
       "      <td>-9.593870</td>\n",
       "      <td>0</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.85272939 3.70408311 4.44153828] Zr, [0.228...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.100232</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.056900</td>\n",
       "      <td>-9.593868</td>\n",
       "      <td>0</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.85291178 3.70391033 4.44185941] Zr, [0.227...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.101241</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.056898</td>\n",
       "      <td>-9.593866</td>\n",
       "      <td>0</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.85307367 3.70349685 4.4420747 ] Zr, [0.227...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.100766</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.056889</td>\n",
       "      <td>-9.593857</td>\n",
       "      <td>0</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.85220447 3.70270231 4.4411749 ] Zr, [0.227...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.100772</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.056889</td>\n",
       "      <td>-9.593857</td>\n",
       "      <td>0</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.85218545 3.70275872 4.44111552] Zr, [0.227...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.100780</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.032976</td>\n",
       "      <td>-9.569944</td>\n",
       "      <td>1</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[1.35623771 1.22996757 0.1635448 ] Zr, [1.356...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.513832</td>\n",
       "      <td>29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.032967</td>\n",
       "      <td>-9.569934</td>\n",
       "      <td>1</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[1.35621643 1.23036135 0.16409659] Zr, [1.356...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.512709</td>\n",
       "      <td>29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.032957</td>\n",
       "      <td>-9.569925</td>\n",
       "      <td>1</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[1.35670995 1.23153948 0.163581  ] Zr, [1.356...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.513512</td>\n",
       "      <td>29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.032949</td>\n",
       "      <td>-9.569917</td>\n",
       "      <td>1</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[1.35719055 1.23255449 0.16429115] Zr, [1.357...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.512753</td>\n",
       "      <td>29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>62</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.032940</td>\n",
       "      <td>-9.569907</td>\n",
       "      <td>1</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[1.35715966 1.23284504 0.16397933] Zr, [1.357...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.513173</td>\n",
       "      <td>29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1161</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.009396</td>\n",
       "      <td>-9.546364</td>\n",
       "      <td>14</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.58180336e+00 2.58182322e+00 5.93759188e-04...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.034036</td>\n",
       "      <td>125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1187</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.009396</td>\n",
       "      <td>-9.546363</td>\n",
       "      <td>14</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.58224575e+00 2.58191568e+00 8.85481727e-04...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.033384</td>\n",
       "      <td>125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1181</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.009392</td>\n",
       "      <td>-9.546360</td>\n",
       "      <td>14</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.58296388e+00 2.58196920e+00 1.34390932e-03...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.030034</td>\n",
       "      <td>125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1180</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.009391</td>\n",
       "      <td>-9.546359</td>\n",
       "      <td>14</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.58218436e+00 2.58184370e+00 8.82030665e-04...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.032999</td>\n",
       "      <td>125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1168</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-8.009374</td>\n",
       "      <td>-9.546341</td>\n",
       "      <td>14</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.58250808e+00 2.58208206e+00 9.65028267e-04...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.032951</td>\n",
       "      <td>125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>927</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.986474</td>\n",
       "      <td>-9.523441</td>\n",
       "      <td>5</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.84233583 0.90468546 2.71490722] Zr, [1.282...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.400476</td>\n",
       "      <td>225</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>926</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.986471</td>\n",
       "      <td>-9.523438</td>\n",
       "      <td>5</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.8425556  0.90467693 2.71430793] Zr, [1.282...</td>\n",
       "      <td>True</td>\n",
       "      <td>2.400426</td>\n",
       "      <td>225</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982412</td>\n",
       "      <td>-9.519380</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.65969157 1.2997801  1.49777842] Zr, [1.371...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.507847</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982373</td>\n",
       "      <td>-9.519341</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.66253792 1.29960545 1.49839971] Zr, [1.370...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.491382</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982335</td>\n",
       "      <td>-9.519303</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.66283827 1.30278042 1.49853554] Zr, [1.370...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.464264</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>65</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982247</td>\n",
       "      <td>-9.519214</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.66157261 1.3074229  1.49870806] Zr, [1.370...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.423580</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>77</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.982104</td>\n",
       "      <td>-9.519071</td>\n",
       "      <td>2</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[3.66078471 1.3143346  1.49821437] Zr, [1.369...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.367775</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>276</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.942265</td>\n",
       "      <td>-9.479232</td>\n",
       "      <td>3</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[0.84375476 1.39417501 5.84686055] Zr, [0.843...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.109196</td>\n",
       "      <td>62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>577</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.942261</td>\n",
       "      <td>-9.479229</td>\n",
       "      <td>3</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[0.84384072 1.3939795  5.84649992] Zr, [0.843...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.111659</td>\n",
       "      <td>62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>482</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.942258</td>\n",
       "      <td>-9.479225</td>\n",
       "      <td>3</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[0.84388549 1.39395452 5.84725204] Zr, [0.843...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.109338</td>\n",
       "      <td>62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>226</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.942257</td>\n",
       "      <td>-9.479225</td>\n",
       "      <td>3</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[0.84379916 1.3940506  5.8460795 ] Zr, [0.843...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.111705</td>\n",
       "      <td>62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>180</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.942256</td>\n",
       "      <td>-9.479224</td>\n",
       "      <td>3</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[0.84366444 1.39396418 5.84895046] Zr, [0.843...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.107523</td>\n",
       "      <td>62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1896</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901468</td>\n",
       "      <td>-9.438435</td>\n",
       "      <td>41</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61389089e+00 0.00000000e+00 1.60054656e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.950326</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>967</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901465</td>\n",
       "      <td>-9.438433</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61330891e+00 2.61330891e+00 3.20038039e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.950561</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1028</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901464</td>\n",
       "      <td>-9.438432</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61336324e+00 2.61336324e+00 3.20044693e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.950136</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1140</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901459</td>\n",
       "      <td>-9.438427</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61376307e+00 2.61376307e+00 3.20093658e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.948152</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1373</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901458</td>\n",
       "      <td>-9.438426</td>\n",
       "      <td>41</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61383554e+00 0.00000000e+00 1.60051266e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.952149</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1219</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901455</td>\n",
       "      <td>-9.438423</td>\n",
       "      <td>41</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61403461e+00 0.00000000e+00 1.60063456e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.952248</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1769</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901446</td>\n",
       "      <td>-9.438413</td>\n",
       "      <td>41</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61301739e+00 0.00000000e+00 1.60001169e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.954058</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1950</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901445</td>\n",
       "      <td>-9.438413</td>\n",
       "      <td>41</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61395205e+00 0.00000000e+00 1.60058401e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.954346</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1059</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901437</td>\n",
       "      <td>-9.438405</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61332090e+00 2.61332090e+00 3.20039507e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.943777</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>996</th>\n",
       "      <td>ZrO2</td>\n",
       "      <td>-7.901436</td>\n",
       "      <td>-9.438404</td>\n",
       "      <td>8</td>\n",
       "      <td>Zr4O8</td>\n",
       "      <td>[[2.61305292e+00 2.61305292e+00 3.20006689e-16...</td>\n",
       "      <td>True</td>\n",
       "      <td>1.944469</td>\n",
       "      <td>205</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     formula_pretty  formation_energy_per_atom  final_energy_per_atom  \\\n",
       "id                                                                      \n",
       "5              ZrO2                  -8.056902              -9.593870   \n",
       "7              ZrO2                  -8.056900              -9.593868   \n",
       "0              ZrO2                  -8.056898              -9.593866   \n",
       "2              ZrO2                  -8.056889              -9.593857   \n",
       "12             ZrO2                  -8.056889              -9.593857   \n",
       "33             ZrO2                  -8.032976              -9.569944   \n",
       "29             ZrO2                  -8.032967              -9.569934   \n",
       "30             ZrO2                  -8.032957              -9.569925   \n",
       "25             ZrO2                  -8.032949              -9.569917   \n",
       "62             ZrO2                  -8.032940              -9.569907   \n",
       "1161           ZrO2                  -8.009396              -9.546364   \n",
       "1187           ZrO2                  -8.009396              -9.546363   \n",
       "1181           ZrO2                  -8.009392              -9.546360   \n",
       "1180           ZrO2                  -8.009391              -9.546359   \n",
       "1168           ZrO2                  -8.009374              -9.546341   \n",
       "927            ZrO2                  -7.986474              -9.523441   \n",
       "926            ZrO2                  -7.986471              -9.523438   \n",
       "78             ZrO2                  -7.982412              -9.519380   \n",
       "63             ZrO2                  -7.982373              -9.519341   \n",
       "69             ZrO2                  -7.982335              -9.519303   \n",
       "65             ZrO2                  -7.982247              -9.519214   \n",
       "77             ZrO2                  -7.982104              -9.519071   \n",
       "276            ZrO2                  -7.942265              -9.479232   \n",
       "577            ZrO2                  -7.942261              -9.479229   \n",
       "482            ZrO2                  -7.942258              -9.479225   \n",
       "226            ZrO2                  -7.942257              -9.479225   \n",
       "180            ZrO2                  -7.942256              -9.479224   \n",
       "1896           ZrO2                  -7.901468              -9.438435   \n",
       "967            ZrO2                  -7.901465              -9.438433   \n",
       "1028           ZrO2                  -7.901464              -9.438432   \n",
       "1140           ZrO2                  -7.901459              -9.438427   \n",
       "1373           ZrO2                  -7.901458              -9.438426   \n",
       "1219           ZrO2                  -7.901455              -9.438423   \n",
       "1769           ZrO2                  -7.901446              -9.438413   \n",
       "1950           ZrO2                  -7.901445              -9.438413   \n",
       "1059           ZrO2                  -7.901437              -9.438405   \n",
       "996            ZrO2                  -7.901436              -9.438404   \n",
       "\n",
       "      structure_group target_formula  \\\n",
       "id                                     \n",
       "5                   0          Zr4O8   \n",
       "7                   0          Zr4O8   \n",
       "0                   0          Zr4O8   \n",
       "2                   0          Zr4O8   \n",
       "12                  0          Zr4O8   \n",
       "33                  1          Zr4O8   \n",
       "29                  1          Zr4O8   \n",
       "30                  1          Zr4O8   \n",
       "25                  1          Zr4O8   \n",
       "62                  1          Zr4O8   \n",
       "1161               14          Zr4O8   \n",
       "1187               14          Zr4O8   \n",
       "1181               14          Zr4O8   \n",
       "1180               14          Zr4O8   \n",
       "1168               14          Zr4O8   \n",
       "927                 5          Zr4O8   \n",
       "926                 5          Zr4O8   \n",
       "78                  2          Zr4O8   \n",
       "63                  2          Zr4O8   \n",
       "69                  2          Zr4O8   \n",
       "65                  2          Zr4O8   \n",
       "77                  2          Zr4O8   \n",
       "276                 3          Zr4O8   \n",
       "577                 3          Zr4O8   \n",
       "482                 3          Zr4O8   \n",
       "226                 3          Zr4O8   \n",
       "180                 3          Zr4O8   \n",
       "1896               41          Zr4O8   \n",
       "967                 8          Zr4O8   \n",
       "1028                8          Zr4O8   \n",
       "1140                8          Zr4O8   \n",
       "1373               41          Zr4O8   \n",
       "1219               41          Zr4O8   \n",
       "1769               41          Zr4O8   \n",
       "1950               41          Zr4O8   \n",
       "1059                8          Zr4O8   \n",
       "996                 8          Zr4O8   \n",
       "\n",
       "                                              structure  converged  \\\n",
       "id                                                                   \n",
       "5     [[2.85272939 3.70408311 4.44153828] Zr, [0.228...       True   \n",
       "7     [[2.85291178 3.70391033 4.44185941] Zr, [0.227...       True   \n",
       "0     [[2.85307367 3.70349685 4.4420747 ] Zr, [0.227...       True   \n",
       "2     [[2.85220447 3.70270231 4.4411749 ] Zr, [0.227...       True   \n",
       "12    [[2.85218545 3.70275872 4.44111552] Zr, [0.227...       True   \n",
       "33    [[1.35623771 1.22996757 0.1635448 ] Zr, [1.356...       True   \n",
       "29    [[1.35621643 1.23036135 0.16409659] Zr, [1.356...       True   \n",
       "30    [[1.35670995 1.23153948 0.163581  ] Zr, [1.356...       True   \n",
       "25    [[1.35719055 1.23255449 0.16429115] Zr, [1.357...       True   \n",
       "62    [[1.35715966 1.23284504 0.16397933] Zr, [1.357...       True   \n",
       "1161  [[2.58180336e+00 2.58182322e+00 5.93759188e-04...       True   \n",
       "1187  [[2.58224575e+00 2.58191568e+00 8.85481727e-04...       True   \n",
       "1181  [[2.58296388e+00 2.58196920e+00 1.34390932e-03...       True   \n",
       "1180  [[2.58218436e+00 2.58184370e+00 8.82030665e-04...       True   \n",
       "1168  [[2.58250808e+00 2.58208206e+00 9.65028267e-04...       True   \n",
       "927   [[3.84233583 0.90468546 2.71490722] Zr, [1.282...       True   \n",
       "926   [[3.8425556  0.90467693 2.71430793] Zr, [1.282...       True   \n",
       "78    [[3.65969157 1.2997801  1.49777842] Zr, [1.371...       True   \n",
       "63    [[3.66253792 1.29960545 1.49839971] Zr, [1.370...       True   \n",
       "69    [[3.66283827 1.30278042 1.49853554] Zr, [1.370...       True   \n",
       "65    [[3.66157261 1.3074229  1.49870806] Zr, [1.370...       True   \n",
       "77    [[3.66078471 1.3143346  1.49821437] Zr, [1.369...       True   \n",
       "276   [[0.84375476 1.39417501 5.84686055] Zr, [0.843...       True   \n",
       "577   [[0.84384072 1.3939795  5.84649992] Zr, [0.843...       True   \n",
       "482   [[0.84388549 1.39395452 5.84725204] Zr, [0.843...       True   \n",
       "226   [[0.84379916 1.3940506  5.8460795 ] Zr, [0.843...       True   \n",
       "180   [[0.84366444 1.39396418 5.84895046] Zr, [0.843...       True   \n",
       "1896  [[2.61389089e+00 0.00000000e+00 1.60054656e-16...       True   \n",
       "967   [[2.61330891e+00 2.61330891e+00 3.20038039e-16...       True   \n",
       "1028  [[2.61336324e+00 2.61336324e+00 3.20044693e-16...       True   \n",
       "1140  [[2.61376307e+00 2.61376307e+00 3.20093658e-16...       True   \n",
       "1373  [[2.61383554e+00 0.00000000e+00 1.60051266e-16...       True   \n",
       "1219  [[2.61403461e+00 0.00000000e+00 1.60063456e-16...       True   \n",
       "1769  [[2.61301739e+00 0.00000000e+00 1.60001169e-16...       True   \n",
       "1950  [[2.61395205e+00 0.00000000e+00 1.60058401e-16...       True   \n",
       "1059  [[2.61332090e+00 2.61332090e+00 3.20039507e-16...       True   \n",
       "996   [[2.61305292e+00 2.61305292e+00 3.20006689e-16...       True   \n",
       "\n",
       "      dissimilarity  space_group_num  \n",
       "id                                    \n",
       "5          0.100232               14  \n",
       "7          0.101241               14  \n",
       "0          0.100766               14  \n",
       "2          0.100772               14  \n",
       "12         0.100780               14  \n",
       "33         0.513832               29  \n",
       "29         0.512709               29  \n",
       "30         0.513512               29  \n",
       "25         0.512753               29  \n",
       "62         0.513173               29  \n",
       "1161       2.034036              125  \n",
       "1187       2.033384              125  \n",
       "1181       2.030034              125  \n",
       "1180       2.032999              125  \n",
       "1168       2.032951              125  \n",
       "927        2.400476              225  \n",
       "926        2.400426              225  \n",
       "78         1.507847               18  \n",
       "63         1.491382               18  \n",
       "69         1.464264               18  \n",
       "65         1.423580               18  \n",
       "77         1.367775               18  \n",
       "276        1.109196               62  \n",
       "577        1.111659               62  \n",
       "482        1.109338               62  \n",
       "226        1.111705               62  \n",
       "180        1.107523               62  \n",
       "1896       1.950326              205  \n",
       "967        1.950561              205  \n",
       "1028       1.950136              205  \n",
       "1140       1.948152              205  \n",
       "1373       1.952149              205  \n",
       "1219       1.952248              205  \n",
       "1769       1.954058              205  \n",
       "1950       1.954346              205  \n",
       "1059       1.943777              205  \n",
       "996        1.944469              205  "
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "final_result = opt_candidate_result.sort_values('formation_energy_per_atom').groupby('structure_group').head(5)\n",
    "final_result = final_result.assign(\n",
    "    dissimilarity=calculate_similarity(true_struct, *final_result.structure, n_jobs=1),\n",
    "    space_group_num=final_result.structure.apply(lambda s: s.get_space_group_info()[1]),\n",
    ")\n",
    "\n",
    "final_result.to_pickle('generated/final_proposed_candidate.pd.xz')\n",
    "final_result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "The proposed structures with id 5, 7, 0, 2, and 12 have the same crystalline structure as the ground truth."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}