notebooks/simple_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: pandas in ./.venv/lib/python3.8/site-packages (1.0.3)\n",
      "Requirement already satisfied: numpy in ./.venv/lib/python3.8/site-packages (1.18.3)\n",
      "Requirement already satisfied: gql in ./.venv/lib/python3.8/site-packages (0.4.0)\n",
      "Requirement already satisfied: pytz>=2017.2 in ./.venv/lib/python3.8/site-packages (from pandas) (2020.1)\n",
      "Requirement already satisfied: python-dateutil>=2.6.1 in ./.venv/lib/python3.8/site-packages (from pandas) (2.8.1)\n",
      "Requirement already satisfied: six>=1.10.0 in ./.venv/lib/python3.8/site-packages (from gql) (1.14.0)\n",
      "Requirement already satisfied: graphql-core<3,>=2 in ./.venv/lib/python3.8/site-packages (from gql) (2.3.1)\n",
      "Requirement already satisfied: promise<3,>=2.0 in ./.venv/lib/python3.8/site-packages (from gql) (2.3)\n",
      "Requirement already satisfied: requests<3,>=2.12 in ./.venv/lib/python3.8/site-packages (from gql) (2.23.0)\n",
      "Requirement already satisfied: rx<2,>=1.6 in ./.venv/lib/python3.8/site-packages (from graphql-core<3,>=2->gql) (1.6.1)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in ./.venv/lib/python3.8/site-packages (from requests<3,>=2.12->gql) (2020.4.5.1)\n",
      "Requirement already satisfied: idna<3,>=2.5 in ./.venv/lib/python3.8/site-packages (from requests<3,>=2.12->gql) (2.9)\n",
      "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in ./.venv/lib/python3.8/site-packages (from requests<3,>=2.12->gql) (1.25.9)\n",
      "Requirement already satisfied: chardet<4,>=3.0.2 in ./.venv/lib/python3.8/site-packages (from requests<3,>=2.12->gql) (3.0.4)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: You are using pip version 19.2.3, however version 20.1 is available.\n",
      "You should consider upgrading via the 'pip install --upgrade pip' command.\n"
     ]
    }
   ],
   "source": [
    "%%bash\n",
    "\n",
    "pip install pandas numpy gql"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gql import gql, Client\n",
    "from gql.transport.requests import RequestsHTTPTransport\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "metadata": {},
   "outputs": [],
   "source": [
    "# display full columns\n",
    "pd.set_option(\"display.max_colwidth\", 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use a local server\n",
    "# GRAPHQL_ENDPOINT = \"http://localhost:8080/v1/graphql\"\n",
    "\n",
    "# Use the public endpoint\n",
    "# !! Please be gentle with your queries! \n",
    "# !! This is running on a small postgres server that I'm paying for, so please don't overload it with automated scripts. \n",
    "# !! As long as you're running queries by hand through notebooks everything should be fine.\n",
    "GRAPHQL_ENDPOINT = \"http://34.107.246.233/v1/graphql\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a graphql client\n",
    "\n",
    "transport = RequestsHTTPTransport(\n",
    "    url=GRAPHQL_ENDPOINT,\n",
    "    use_json=True,\n",
    "    headers={\"Content-type\": \"application/json\"},\n",
    "    verify=False)\n",
    "\n",
    "client = Client(\n",
    "    retries=3,\n",
    "    transport=transport,\n",
    "    fetch_schema_from_transport=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Query the database\n",
    "\n",
    "Let's get some data. The following code queries the GraphQL endpoint and gets the 2-level citation graph for a specific paper. It retrieves all papers at most 2 hops away. Assuming each paper on average cites 30 other papers, the result would be 1 + 30 * 30 = ~900 records. Note that the database currently only contains papers labeled as \"Computer Science\", so cited papers from other fields will not show up."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Paper title is matched via sql LIKE, so use can use % operators for text matching\n",
    "PAPER_TITLE = \"%Mastering Atari, Go, Chess and Shogi by Planning%\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "metadata": {},
   "outputs": [],
   "source": [
    "query = gql(\"\"\"\n",
    "fragment paper_fields on papers {\n",
    "  id\n",
    "  title\n",
    "  year\n",
    "  doi_url\n",
    "  s2_url\n",
    "  num_citations\n",
    "}\n",
    "\n",
    "query papers($title: String!) {\n",
    "  papers(limit: 1, where: {title: {_like: $title}}, offset: 0) {\n",
    "    ...paper_fields\n",
    "    cites(args: {limit_: 100}) {\n",
    "      ...paper_fields\n",
    "      cites(args: {limit_: 100}) {\n",
    "        ...paper_fields\n",
    "      }\n",
    "    }\n",
    "  }\n",
    "}\n",
    "\"\"\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "metadata": {},
   "outputs": [],
   "source": [
    "papers = client.execute(query, variable_values={\"title\": PAPER_TITLE})\n",
    "root = papers[\"papers\"][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [],
   "source": [
    "def process_recursively(paper, fn):\n",
    "    \"\"\"Helper function to process the recursive graph data structure. It runs a function for each element in the graph\"\"\"\n",
    "    fn(paper)\n",
    "    if not \"cites\" in paper:\n",
    "        return\n",
    "    for cited_paper in paper[\"cites\"]:\n",
    "        process_recursively(cited_paper, fn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create DataFrame\n",
    "\n",
    "Next, let's create a dataframe of all papers in the graph. Here, duplicates are eliminated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "744 unique papers\n"
     ]
    },
    {
     "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>title</th>\n",
       "      <th>year</th>\n",
       "      <th>doi_url</th>\n",
       "      <th>s2_url</th>\n",
       "      <th>num_citations</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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>c39fb7a46335c23f7529dd6f9f980462fd38653a</th>\n",
       "      <td>Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model</td>\n",
       "      <td>2019</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/c39fb7a46335c23f7529dd6f9f980462fd38653a</td>\n",
       "      <td>25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d</th>\n",
       "      <td>Between MDPs and Semi-MDPs: A Framework for Temporal Abstraction in Reinforcement Learning</td>\n",
       "      <td>1999</td>\n",
       "      <td>https://doi.org/10.1016/S0004-3702%2899%2900052-1</td>\n",
       "      <td>https://semanticscholar.org/paper/0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d</td>\n",
       "      <td>1622</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>036373f17e5e47bcadc289e6c57d61cf5e08fe3d</th>\n",
       "      <td>Hierarchical Solution of Markov Decision Processes using Macro-actions</td>\n",
       "      <td>1998</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/036373f17e5e47bcadc289e6c57d61cf5e08fe3d</td>\n",
       "      <td>228</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>07b6e294c47ef0d72b3229ca6b891dd772adb47d</th>\n",
       "      <td>Theoretical Results on Reinforcement Learning with Temporally Abstract Options</td>\n",
       "      <td>1998</td>\n",
       "      <td>https://doi.org/10.1007/BFb0026709</td>\n",
       "      <td>https://semanticscholar.org/paper/07b6e294c47ef0d72b3229ca6b891dd772adb47d</td>\n",
       "      <td>74</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0a5bac1a42c05d4711bcd23c8caae60eb886fbb3</th>\n",
       "      <td>Planning under Time Constraints in Stochastic Domains</td>\n",
       "      <td>1995</td>\n",
       "      <td>https://doi.org/10.1016/0004-3702%2894%2900086-G</td>\n",
       "      <td>https://semanticscholar.org/paper/0a5bac1a42c05d4711bcd23c8caae60eb886fbb3</td>\n",
       "      <td>214</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                                               title  \\\n",
       "id                                                                                                                                     \n",
       "c39fb7a46335c23f7529dd6f9f980462fd38653a  Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model                        \n",
       "0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d  Between MDPs and Semi-MDPs: A Framework for Temporal Abstraction in Reinforcement Learning   \n",
       "036373f17e5e47bcadc289e6c57d61cf5e08fe3d  Hierarchical Solution of Markov Decision Processes using Macro-actions                       \n",
       "07b6e294c47ef0d72b3229ca6b891dd772adb47d  Theoretical Results on Reinforcement Learning with Temporally Abstract Options               \n",
       "0a5bac1a42c05d4711bcd23c8caae60eb886fbb3  Planning under Time Constraints in Stochastic Domains                                        \n",
       "\n",
       "                                          year  \\\n",
       "id                                               \n",
       "c39fb7a46335c23f7529dd6f9f980462fd38653a  2019   \n",
       "0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d  1999   \n",
       "036373f17e5e47bcadc289e6c57d61cf5e08fe3d  1998   \n",
       "07b6e294c47ef0d72b3229ca6b891dd772adb47d  1998   \n",
       "0a5bac1a42c05d4711bcd23c8caae60eb886fbb3  1995   \n",
       "\n",
       "                                                                                    doi_url  \\\n",
       "id                                                                                            \n",
       "c39fb7a46335c23f7529dd6f9f980462fd38653a                                                      \n",
       "0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d  https://doi.org/10.1016/S0004-3702%2899%2900052-1   \n",
       "036373f17e5e47bcadc289e6c57d61cf5e08fe3d                                                      \n",
       "07b6e294c47ef0d72b3229ca6b891dd772adb47d  https://doi.org/10.1007/BFb0026709                  \n",
       "0a5bac1a42c05d4711bcd23c8caae60eb886fbb3  https://doi.org/10.1016/0004-3702%2894%2900086-G    \n",
       "\n",
       "                                                                                                              s2_url  \\\n",
       "id                                                                                                                     \n",
       "c39fb7a46335c23f7529dd6f9f980462fd38653a  https://semanticscholar.org/paper/c39fb7a46335c23f7529dd6f9f980462fd38653a   \n",
       "0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d  https://semanticscholar.org/paper/0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d   \n",
       "036373f17e5e47bcadc289e6c57d61cf5e08fe3d  https://semanticscholar.org/paper/036373f17e5e47bcadc289e6c57d61cf5e08fe3d   \n",
       "07b6e294c47ef0d72b3229ca6b891dd772adb47d  https://semanticscholar.org/paper/07b6e294c47ef0d72b3229ca6b891dd772adb47d   \n",
       "0a5bac1a42c05d4711bcd23c8caae60eb886fbb3  https://semanticscholar.org/paper/0a5bac1a42c05d4711bcd23c8caae60eb886fbb3   \n",
       "\n",
       "                                          num_citations  \n",
       "id                                                       \n",
       "c39fb7a46335c23f7529dd6f9f980462fd38653a  25             \n",
       "0e7638dc16a5e5e9e46c91272bfb9c3dd242ef6d  1622           \n",
       "036373f17e5e47bcadc289e6c57d61cf5e08fe3d  228            \n",
       "07b6e294c47ef0d72b3229ca6b891dd772adb47d  74             \n",
       "0a5bac1a42c05d4711bcd23c8caae60eb886fbb3  214            "
      ]
     },
     "execution_count": 226,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Create a DataFrame of all papers that appear in this subgraph\n",
    "paper_map = {}\n",
    "def add_paper(paper):\n",
    "    fields = { k : paper[k] for k in paper if k != \"cites\" }\n",
    "    paper_map[paper[\"id\"]] = fields\n",
    "process_recursively(root, add_paper)\n",
    "papers_df = pd.DataFrame.from_records(list(paper_map.values()), index=\"id\")\n",
    "\n",
    "print(f\"{len(papers_df)} unique papers\")\n",
    "papers_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Most Popular Papers\n",
    "\n",
    "The following shows papers with the overall most citations present in this subgraph. These are the popular papers that somewhat relevant to the root paper."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {},
   "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>title</th>\n",
       "      <th>year</th>\n",
       "      <th>doi_url</th>\n",
       "      <th>s2_url</th>\n",
       "      <th>num_citations</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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>abd1c342495432171beb7ca8fd9551ef13cbd0ff</th>\n",
       "      <td>ImageNet Classification with Deep Convolutional Neural Networks</td>\n",
       "      <td>2012</td>\n",
       "      <td>https://doi.org/10.1061/%28ASCE%29GT.1943-5606.0001284</td>\n",
       "      <td>https://semanticscholar.org/paper/abd1c342495432171beb7ca8fd9551ef13cbd0ff</td>\n",
       "      <td>40246</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2c03df8b48bf3fa39054345bafabfeff15bfd11d</th>\n",
       "      <td>Deep Residual Learning for Image Recognition</td>\n",
       "      <td>2016</td>\n",
       "      <td>https://doi.org/10.1109/cvpr.2016.90</td>\n",
       "      <td>https://semanticscholar.org/paper/2c03df8b48bf3fa39054345bafabfeff15bfd11d</td>\n",
       "      <td>32655</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986</th>\n",
       "      <td>Random Forests</td>\n",
       "      <td>2004</td>\n",
       "      <td>https://doi.org/10.1023/A%3A1010933404324</td>\n",
       "      <td>https://semanticscholar.org/paper/13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986</td>\n",
       "      <td>24563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44d2abe2175df8153f465f6c39b68b76a0d40ab9</th>\n",
       "      <td>Long Short-Term Memory</td>\n",
       "      <td>1997</td>\n",
       "      <td>https://doi.org/10.1162/neco.1997.9.8.1735</td>\n",
       "      <td>https://semanticscholar.org/paper/44d2abe2175df8153f465f6c39b68b76a0d40ab9</td>\n",
       "      <td>20985</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97efafdb4a3942ab3efba53ded7413199f79c054</th>\n",
       "      <td>Reinforcement Learning: An Introduction</td>\n",
       "      <td>2005</td>\n",
       "      <td>https://doi.org/10.1109/TNN.1998.712192</td>\n",
       "      <td>https://semanticscholar.org/paper/97efafdb4a3942ab3efba53ded7413199f79c054</td>\n",
       "      <td>18170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4f607f03272e4d62708f5b2441355f9e005cb452</th>\n",
       "      <td>Convex Optimization</td>\n",
       "      <td>2006</td>\n",
       "      <td>https://doi.org/10.1017/CBO9780511804441</td>\n",
       "      <td>https://semanticscholar.org/paper/4f607f03272e4d62708f5b2441355f9e005cb452</td>\n",
       "      <td>16074</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2e62d1345b340d5fda3b092c460264b9543bc4b5</th>\n",
       "      <td>Genetic Algorithms in Search Optimization and Machine Learning</td>\n",
       "      <td>1989</td>\n",
       "      <td>https://doi.org/10.5860/choice.27-0936</td>\n",
       "      <td>https://semanticscholar.org/paper/2e62d1345b340d5fda3b092c460264b9543bc4b5</td>\n",
       "      <td>16067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4b4279db68b16e20fbc56f9d41980a950191d30a</th>\n",
       "      <td>Adaptation in natural and artificial systems</td>\n",
       "      <td>1975</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/4b4279db68b16e20fbc56f9d41980a950191d30a</td>\n",
       "      <td>15985</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>162d958ff885f1462aeda91cd72582323fd6a1f4</th>\n",
       "      <td>Gradient-based learning applied to document recognition</td>\n",
       "      <td>1998</td>\n",
       "      <td>https://doi.org/10.1109/5.726791</td>\n",
       "      <td>https://semanticscholar.org/paper/162d958ff885f1462aeda91cd72582323fd6a1f4</td>\n",
       "      <td>15423</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>e15cf50aa89fee8535703b9f9512fca5bfc43327</th>\n",
       "      <td>Going deeper with convolutions</td>\n",
       "      <td>2015</td>\n",
       "      <td>https://doi.org/10.1109/CVPR.2015.7298594</td>\n",
       "      <td>https://semanticscholar.org/paper/e15cf50aa89fee8535703b9f9512fca5bfc43327</td>\n",
       "      <td>13348</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4d376d6978dad0374edfa6709c9556b42d3594d3</th>\n",
       "      <td>Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift</td>\n",
       "      <td>2015</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/4d376d6978dad0374edfa6709c9556b42d3594d3</td>\n",
       "      <td>13232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54e325aee6b2d476bbbb88615ac15e251c6e8214</th>\n",
       "      <td>Generative Adversarial Nets</td>\n",
       "      <td>2014</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/54e325aee6b2d476bbbb88615ac15e251c6e8214</td>\n",
       "      <td>11910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>e74f9b7f8eec6ba4704c206b93bc8079af3da4bd</th>\n",
       "      <td>ImageNet Large Scale Visual Recognition Challenge</td>\n",
       "      <td>2015</td>\n",
       "      <td>https://doi.org/10.1007/s11263-015-0816-y</td>\n",
       "      <td>https://semanticscholar.org/paper/e74f9b7f8eec6ba4704c206b93bc8079af3da4bd</td>\n",
       "      <td>10959</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>d2c733e34d48784a37d717fe43d9e93277a8c53e</th>\n",
       "      <td>ImageNet: A large-scale hierarchical image database</td>\n",
       "      <td>2009</td>\n",
       "      <td>https://doi.org/10.1109/CVPRW.2009.5206848</td>\n",
       "      <td>https://semanticscholar.org/paper/d2c733e34d48784a37d717fe43d9e93277a8c53e</td>\n",
       "      <td>10725</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6a17ebeeb80cd696bc83a288f1a77ddfc1467079</th>\n",
       "      <td>System Identification: Theory for the User</td>\n",
       "      <td>1987</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/6a17ebeeb80cd696bc83a288f1a77ddfc1467079</td>\n",
       "      <td>9534</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5</th>\n",
       "      <td>Neural Machine Translation by Jointly Learning to Align and Translate</td>\n",
       "      <td>2014</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5</td>\n",
       "      <td>9506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>111fd833a4ae576cfdbb27d87d2f8fc0640af355</th>\n",
       "      <td>Learning internal representations by error propagation</td>\n",
       "      <td>1986</td>\n",
       "      <td>https://doi.org/10.1016/B978-1-4832-1446-7.50035-2</td>\n",
       "      <td>https://semanticscholar.org/paper/111fd833a4ae576cfdbb27d87d2f8fc0640af355</td>\n",
       "      <td>9234</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>82266f6103bade9005ec555ed06ba20b5210ff22</th>\n",
       "      <td>Gaussian processes for machine learning</td>\n",
       "      <td>2006</td>\n",
       "      <td>https://doi.org/10.7551/mitpress%2F3206.001.0001</td>\n",
       "      <td>https://semanticscholar.org/paper/82266f6103bade9005ec555ed06ba20b5210ff22</td>\n",
       "      <td>8025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e</th>\n",
       "      <td>Reducing the dimensionality of data with neural networks.</td>\n",
       "      <td>2006</td>\n",
       "      <td>https://doi.org/10.1126/science.1127647</td>\n",
       "      <td>https://semanticscholar.org/paper/46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e</td>\n",
       "      <td>7590</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>b07ce649d6f6eb636872527104b0209d3edc8188</th>\n",
       "      <td>Pattern classification and scene analysis</td>\n",
       "      <td>1973</td>\n",
       "      <td>https://doi.org/10.2307/2286028</td>\n",
       "      <td>https://semanticscholar.org/paper/b07ce649d6f6eb636872527104b0209d3edc8188</td>\n",
       "      <td>7391</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71b7178df5d2b112d07e45038cb5637208659ff7</th>\n",
       "      <td>Microsoft COCO: Common Objects in Context</td>\n",
       "      <td>2014</td>\n",
       "      <td>https://doi.org/10.1007/978-3-319-10602-1_48</td>\n",
       "      <td>https://semanticscholar.org/paper/71b7178df5d2b112d07e45038cb5637208659ff7</td>\n",
       "      <td>6938</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d</th>\n",
       "      <td>Human-level control through deep reinforcement learning</td>\n",
       "      <td>2015</td>\n",
       "      <td>https://doi.org/10.1038/nature14236</td>\n",
       "      <td>https://semanticscholar.org/paper/e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d</td>\n",
       "      <td>6647</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0b544dfe355a5070b60986319a3f51fb45d1348e</th>\n",
       "      <td>Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation</td>\n",
       "      <td>2014</td>\n",
       "      <td>https://doi.org/10.3115/v1%2FD14-1179</td>\n",
       "      <td>https://semanticscholar.org/paper/0b544dfe355a5070b60986319a3f51fb45d1348e</td>\n",
       "      <td>6619</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d</th>\n",
       "      <td>TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems</td>\n",
       "      <td>2015</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d</td>\n",
       "      <td>6558</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>a538b05ebb01a40323997629e171c91aa28b8e2f</th>\n",
       "      <td>Rectified Linear Units Improve Restricted Boltzmann Machines</td>\n",
       "      <td>2010</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/a538b05ebb01a40323997629e171c91aa28b8e2f</td>\n",
       "      <td>6310</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                                                  title  \\\n",
       "id                                                                                                                                        \n",
       "abd1c342495432171beb7ca8fd9551ef13cbd0ff  ImageNet Classification with Deep Convolutional Neural Networks                                 \n",
       "2c03df8b48bf3fa39054345bafabfeff15bfd11d  Deep Residual Learning for Image Recognition                                                    \n",
       "13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986  Random Forests                                                                                  \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  Long Short-Term Memory                                                                          \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  Reinforcement Learning: An Introduction                                                         \n",
       "4f607f03272e4d62708f5b2441355f9e005cb452  Convex Optimization                                                                             \n",
       "2e62d1345b340d5fda3b092c460264b9543bc4b5  Genetic Algorithms in Search Optimization and Machine Learning                                  \n",
       "4b4279db68b16e20fbc56f9d41980a950191d30a  Adaptation in natural and artificial systems                                                    \n",
       "162d958ff885f1462aeda91cd72582323fd6a1f4  Gradient-based learning applied to document recognition                                         \n",
       "e15cf50aa89fee8535703b9f9512fca5bfc43327  Going deeper with convolutions                                                                  \n",
       "4d376d6978dad0374edfa6709c9556b42d3594d3  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift    \n",
       "54e325aee6b2d476bbbb88615ac15e251c6e8214  Generative Adversarial Nets                                                                     \n",
       "e74f9b7f8eec6ba4704c206b93bc8079af3da4bd  ImageNet Large Scale Visual Recognition Challenge                                               \n",
       "d2c733e34d48784a37d717fe43d9e93277a8c53e  ImageNet: A large-scale hierarchical image database                                             \n",
       "6a17ebeeb80cd696bc83a288f1a77ddfc1467079  System Identification: Theory for the User                                                      \n",
       "fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5  Neural Machine Translation by Jointly Learning to Align and Translate                           \n",
       "111fd833a4ae576cfdbb27d87d2f8fc0640af355  Learning internal representations by error propagation                                          \n",
       "82266f6103bade9005ec555ed06ba20b5210ff22  Gaussian processes for machine learning                                                         \n",
       "46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e  Reducing the dimensionality of data with neural networks.                                       \n",
       "b07ce649d6f6eb636872527104b0209d3edc8188  Pattern classification and scene analysis                                                       \n",
       "71b7178df5d2b112d07e45038cb5637208659ff7  Microsoft COCO: Common Objects in Context                                                       \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  Human-level control through deep reinforcement learning                                         \n",
       "0b544dfe355a5070b60986319a3f51fb45d1348e  Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation   \n",
       "9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d  TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems                   \n",
       "a538b05ebb01a40323997629e171c91aa28b8e2f  Rectified Linear Units Improve Restricted Boltzmann Machines                                    \n",
       "\n",
       "                                          year  \\\n",
       "id                                               \n",
       "abd1c342495432171beb7ca8fd9551ef13cbd0ff  2012   \n",
       "2c03df8b48bf3fa39054345bafabfeff15bfd11d  2016   \n",
       "13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986  2004   \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  1997   \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  2005   \n",
       "4f607f03272e4d62708f5b2441355f9e005cb452  2006   \n",
       "2e62d1345b340d5fda3b092c460264b9543bc4b5  1989   \n",
       "4b4279db68b16e20fbc56f9d41980a950191d30a  1975   \n",
       "162d958ff885f1462aeda91cd72582323fd6a1f4  1998   \n",
       "e15cf50aa89fee8535703b9f9512fca5bfc43327  2015   \n",
       "4d376d6978dad0374edfa6709c9556b42d3594d3  2015   \n",
       "54e325aee6b2d476bbbb88615ac15e251c6e8214  2014   \n",
       "e74f9b7f8eec6ba4704c206b93bc8079af3da4bd  2015   \n",
       "d2c733e34d48784a37d717fe43d9e93277a8c53e  2009   \n",
       "6a17ebeeb80cd696bc83a288f1a77ddfc1467079  1987   \n",
       "fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5  2014   \n",
       "111fd833a4ae576cfdbb27d87d2f8fc0640af355  1986   \n",
       "82266f6103bade9005ec555ed06ba20b5210ff22  2006   \n",
       "46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e  2006   \n",
       "b07ce649d6f6eb636872527104b0209d3edc8188  1973   \n",
       "71b7178df5d2b112d07e45038cb5637208659ff7  2014   \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  2015   \n",
       "0b544dfe355a5070b60986319a3f51fb45d1348e  2014   \n",
       "9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d  2015   \n",
       "a538b05ebb01a40323997629e171c91aa28b8e2f  2010   \n",
       "\n",
       "                                                                                         doi_url  \\\n",
       "id                                                                                                 \n",
       "abd1c342495432171beb7ca8fd9551ef13cbd0ff  https://doi.org/10.1061/%28ASCE%29GT.1943-5606.0001284   \n",
       "2c03df8b48bf3fa39054345bafabfeff15bfd11d  https://doi.org/10.1109/cvpr.2016.90                     \n",
       "13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986  https://doi.org/10.1023/A%3A1010933404324                \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  https://doi.org/10.1162/neco.1997.9.8.1735               \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  https://doi.org/10.1109/TNN.1998.712192                  \n",
       "4f607f03272e4d62708f5b2441355f9e005cb452  https://doi.org/10.1017/CBO9780511804441                 \n",
       "2e62d1345b340d5fda3b092c460264b9543bc4b5  https://doi.org/10.5860/choice.27-0936                   \n",
       "4b4279db68b16e20fbc56f9d41980a950191d30a                                                           \n",
       "162d958ff885f1462aeda91cd72582323fd6a1f4  https://doi.org/10.1109/5.726791                         \n",
       "e15cf50aa89fee8535703b9f9512fca5bfc43327  https://doi.org/10.1109/CVPR.2015.7298594                \n",
       "4d376d6978dad0374edfa6709c9556b42d3594d3                                                           \n",
       "54e325aee6b2d476bbbb88615ac15e251c6e8214                                                           \n",
       "e74f9b7f8eec6ba4704c206b93bc8079af3da4bd  https://doi.org/10.1007/s11263-015-0816-y                \n",
       "d2c733e34d48784a37d717fe43d9e93277a8c53e  https://doi.org/10.1109/CVPRW.2009.5206848               \n",
       "6a17ebeeb80cd696bc83a288f1a77ddfc1467079                                                           \n",
       "fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5                                                           \n",
       "111fd833a4ae576cfdbb27d87d2f8fc0640af355  https://doi.org/10.1016/B978-1-4832-1446-7.50035-2       \n",
       "82266f6103bade9005ec555ed06ba20b5210ff22  https://doi.org/10.7551/mitpress%2F3206.001.0001         \n",
       "46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e  https://doi.org/10.1126/science.1127647                  \n",
       "b07ce649d6f6eb636872527104b0209d3edc8188  https://doi.org/10.2307/2286028                          \n",
       "71b7178df5d2b112d07e45038cb5637208659ff7  https://doi.org/10.1007/978-3-319-10602-1_48             \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  https://doi.org/10.1038/nature14236                      \n",
       "0b544dfe355a5070b60986319a3f51fb45d1348e  https://doi.org/10.3115/v1%2FD14-1179                    \n",
       "9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d                                                           \n",
       "a538b05ebb01a40323997629e171c91aa28b8e2f                                                           \n",
       "\n",
       "                                                                                                              s2_url  \\\n",
       "id                                                                                                                     \n",
       "abd1c342495432171beb7ca8fd9551ef13cbd0ff  https://semanticscholar.org/paper/abd1c342495432171beb7ca8fd9551ef13cbd0ff   \n",
       "2c03df8b48bf3fa39054345bafabfeff15bfd11d  https://semanticscholar.org/paper/2c03df8b48bf3fa39054345bafabfeff15bfd11d   \n",
       "13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986  https://semanticscholar.org/paper/13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986   \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  https://semanticscholar.org/paper/44d2abe2175df8153f465f6c39b68b76a0d40ab9   \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  https://semanticscholar.org/paper/97efafdb4a3942ab3efba53ded7413199f79c054   \n",
       "4f607f03272e4d62708f5b2441355f9e005cb452  https://semanticscholar.org/paper/4f607f03272e4d62708f5b2441355f9e005cb452   \n",
       "2e62d1345b340d5fda3b092c460264b9543bc4b5  https://semanticscholar.org/paper/2e62d1345b340d5fda3b092c460264b9543bc4b5   \n",
       "4b4279db68b16e20fbc56f9d41980a950191d30a  https://semanticscholar.org/paper/4b4279db68b16e20fbc56f9d41980a950191d30a   \n",
       "162d958ff885f1462aeda91cd72582323fd6a1f4  https://semanticscholar.org/paper/162d958ff885f1462aeda91cd72582323fd6a1f4   \n",
       "e15cf50aa89fee8535703b9f9512fca5bfc43327  https://semanticscholar.org/paper/e15cf50aa89fee8535703b9f9512fca5bfc43327   \n",
       "4d376d6978dad0374edfa6709c9556b42d3594d3  https://semanticscholar.org/paper/4d376d6978dad0374edfa6709c9556b42d3594d3   \n",
       "54e325aee6b2d476bbbb88615ac15e251c6e8214  https://semanticscholar.org/paper/54e325aee6b2d476bbbb88615ac15e251c6e8214   \n",
       "e74f9b7f8eec6ba4704c206b93bc8079af3da4bd  https://semanticscholar.org/paper/e74f9b7f8eec6ba4704c206b93bc8079af3da4bd   \n",
       "d2c733e34d48784a37d717fe43d9e93277a8c53e  https://semanticscholar.org/paper/d2c733e34d48784a37d717fe43d9e93277a8c53e   \n",
       "6a17ebeeb80cd696bc83a288f1a77ddfc1467079  https://semanticscholar.org/paper/6a17ebeeb80cd696bc83a288f1a77ddfc1467079   \n",
       "fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5  https://semanticscholar.org/paper/fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5   \n",
       "111fd833a4ae576cfdbb27d87d2f8fc0640af355  https://semanticscholar.org/paper/111fd833a4ae576cfdbb27d87d2f8fc0640af355   \n",
       "82266f6103bade9005ec555ed06ba20b5210ff22  https://semanticscholar.org/paper/82266f6103bade9005ec555ed06ba20b5210ff22   \n",
       "46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e  https://semanticscholar.org/paper/46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e   \n",
       "b07ce649d6f6eb636872527104b0209d3edc8188  https://semanticscholar.org/paper/b07ce649d6f6eb636872527104b0209d3edc8188   \n",
       "71b7178df5d2b112d07e45038cb5637208659ff7  https://semanticscholar.org/paper/71b7178df5d2b112d07e45038cb5637208659ff7   \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  https://semanticscholar.org/paper/e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d   \n",
       "0b544dfe355a5070b60986319a3f51fb45d1348e  https://semanticscholar.org/paper/0b544dfe355a5070b60986319a3f51fb45d1348e   \n",
       "9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d  https://semanticscholar.org/paper/9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d   \n",
       "a538b05ebb01a40323997629e171c91aa28b8e2f  https://semanticscholar.org/paper/a538b05ebb01a40323997629e171c91aa28b8e2f   \n",
       "\n",
       "                                          num_citations  \n",
       "id                                                       \n",
       "abd1c342495432171beb7ca8fd9551ef13cbd0ff  40246          \n",
       "2c03df8b48bf3fa39054345bafabfeff15bfd11d  32655          \n",
       "13d4c2f76a7c1a4d0a71204e1d5d263a3f5a7986  24563          \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  20985          \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  18170          \n",
       "4f607f03272e4d62708f5b2441355f9e005cb452  16074          \n",
       "2e62d1345b340d5fda3b092c460264b9543bc4b5  16067          \n",
       "4b4279db68b16e20fbc56f9d41980a950191d30a  15985          \n",
       "162d958ff885f1462aeda91cd72582323fd6a1f4  15423          \n",
       "e15cf50aa89fee8535703b9f9512fca5bfc43327  13348          \n",
       "4d376d6978dad0374edfa6709c9556b42d3594d3  13232          \n",
       "54e325aee6b2d476bbbb88615ac15e251c6e8214  11910          \n",
       "e74f9b7f8eec6ba4704c206b93bc8079af3da4bd  10959          \n",
       "d2c733e34d48784a37d717fe43d9e93277a8c53e  10725          \n",
       "6a17ebeeb80cd696bc83a288f1a77ddfc1467079  9534           \n",
       "fa72afa9b2cbc8f0d7b05d52548906610ffbb9c5  9506           \n",
       "111fd833a4ae576cfdbb27d87d2f8fc0640af355  9234           \n",
       "82266f6103bade9005ec555ed06ba20b5210ff22  8025           \n",
       "46eb79e5eec8a4e2b2f5652b66441e8a4c921c3e  7590           \n",
       "b07ce649d6f6eb636872527104b0209d3edc8188  7391           \n",
       "71b7178df5d2b112d07e45038cb5637208659ff7  6938           \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  6647           \n",
       "0b544dfe355a5070b60986319a3f51fb45d1348e  6619           \n",
       "9c9d7247f8c51ec5a02b0d911d1d7b9e8160495d  6558           \n",
       "a538b05ebb01a40323997629e171c91aa28b8e2f  6310           "
      ]
     },
     "execution_count": 227,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "papers_df.sort_values(\"num_citations\", ascending=False).head(25)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Most cited papers within the subgraph\n",
    "\n",
    "The following finds papers that are cited most often **within this subgraph**. These papers tend to be more relevant to the root paper since they are often cited by related papers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {},
   "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>title</th>\n",
       "      <th>year</th>\n",
       "      <th>doi_url</th>\n",
       "      <th>s2_url</th>\n",
       "      <th>num_citations</th>\n",
       "      <th>subgraph_citation_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d</th>\n",
       "      <td>Human-level control through deep reinforcement learning</td>\n",
       "      <td>2015</td>\n",
       "      <td>https://doi.org/10.1038/nature14236</td>\n",
       "      <td>https://semanticscholar.org/paper/e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d</td>\n",
       "      <td>6647</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97efafdb4a3942ab3efba53ded7413199f79c054</th>\n",
       "      <td>Reinforcement Learning: An Introduction</td>\n",
       "      <td>2005</td>\n",
       "      <td>https://doi.org/10.1109/TNN.1998.712192</td>\n",
       "      <td>https://semanticscholar.org/paper/97efafdb4a3942ab3efba53ded7413199f79c054</td>\n",
       "      <td>18170</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>f82e4ff4f003581330338aaae71f60316e58dd26</th>\n",
       "      <td>The Arcade Learning Environment: An Evaluation Platform for General Agents (Extended Abstract)</td>\n",
       "      <td>2013</td>\n",
       "      <td>https://doi.org/10.1613/jair.3912</td>\n",
       "      <td>https://semanticscholar.org/paper/f82e4ff4f003581330338aaae71f60316e58dd26</td>\n",
       "      <td>1003</td>\n",
       "      <td>12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69e76e16740ed69f4dc55361a3d319ac2f1293dd</th>\n",
       "      <td>Asynchronous Methods for Deep Reinforcement Learning</td>\n",
       "      <td>2016</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/69e76e16740ed69f4dc55361a3d319ac2f1293dd</td>\n",
       "      <td>2365</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>e4257bc131c36504a04382290cbc27ca8bb27813</th>\n",
       "      <td>Action-Conditional Video Prediction using Deep Networks in Atari Games</td>\n",
       "      <td>2015</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/e4257bc131c36504a04382290cbc27ca8bb27813</td>\n",
       "      <td>444</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>c6170fa90d3b2efede5a2e1660cb23e1c824f2ca</th>\n",
       "      <td>Prioritized Experience Replay</td>\n",
       "      <td>2015</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/c6170fa90d3b2efede5a2e1660cb23e1c824f2ca</td>\n",
       "      <td>864</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60b7d47758a71978e74edff6dd8dea4d9c791d7a</th>\n",
       "      <td>PILCO: A Model-Based and Data-Efficient Approach to Policy Search</td>\n",
       "      <td>2011</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/60b7d47758a71978e74edff6dd8dea4d9c791d7a</td>\n",
       "      <td>642</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3b9732bb07dc99bde5e1f9f75251c6ea5039373e</th>\n",
       "      <td>Deep Reinforcement Learning with Double Q-Learning</td>\n",
       "      <td>2016</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/3b9732bb07dc99bde5e1f9f75251c6ea5039373e</td>\n",
       "      <td>1388</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54c4cf3a8168c1b70f91cf78a3dc98b671935492</th>\n",
       "      <td>Reinforcement learning for robots using neural networks</td>\n",
       "      <td>1992</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/54c4cf3a8168c1b70f91cf78a3dc98b671935492</td>\n",
       "      <td>529</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>846aedd869a00c09b40f1f1f35673cb22bc87490</th>\n",
       "      <td>Mastering the game of Go with deep neural networks and tree search</td>\n",
       "      <td>2016</td>\n",
       "      <td>https://doi.org/10.1038/nature16961</td>\n",
       "      <td>https://semanticscholar.org/paper/846aedd869a00c09b40f1f1f35673cb22bc87490</td>\n",
       "      <td>4912</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>e635d81a617d1239232a9c9a11a196c53dab8240</th>\n",
       "      <td>Bandit Based Monte-Carlo Planning</td>\n",
       "      <td>2006</td>\n",
       "      <td>https://doi.org/10.1007/11871842_29</td>\n",
       "      <td>https://semanticscholar.org/paper/e635d81a617d1239232a9c9a11a196c53dab8240</td>\n",
       "      <td>1593</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2319a491378867c7049b3da055c5df60e1671158</th>\n",
       "      <td>Playing Atari with Deep Reinforcement Learning</td>\n",
       "      <td>2013</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/2319a491378867c7049b3da055c5df60e1671158</td>\n",
       "      <td>2793</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6ce57ab17fcd507b856a79874063b59555c76b3a</th>\n",
       "      <td>Learning to Predict by the Methods of Temporal Differences</td>\n",
       "      <td>2005</td>\n",
       "      <td>https://doi.org/10.1023/A%3A1022633531479</td>\n",
       "      <td>https://semanticscholar.org/paper/6ce57ab17fcd507b856a79874063b59555c76b3a</td>\n",
       "      <td>2365</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4c05d7caa357148f0bbd61720bdd35f0bc05eb81</th>\n",
       "      <td>Dueling Network Architectures for Deep Reinforcement Learning</td>\n",
       "      <td>2016</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/4c05d7caa357148f0bbd61720bdd35f0bc05eb81</td>\n",
       "      <td>838</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>282001869bd502c7917db8b32b75593addfbbc68</th>\n",
       "      <td>Neural Fitted Q Iteration - First Experiences with a Data Efficient Neural Reinforcement Learning Method</td>\n",
       "      <td>2005</td>\n",
       "      <td>https://doi.org/10.1007/11564096_32</td>\n",
       "      <td>https://semanticscholar.org/paper/282001869bd502c7917db8b32b75593addfbbc68</td>\n",
       "      <td>557</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>d7bd6e3addd8bc8e2e154048300eea15f030ed33</th>\n",
       "      <td>Reinforcement Learning with Unsupervised Auxiliary Tasks</td>\n",
       "      <td>2017</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/d7bd6e3addd8bc8e2e154048300eea15f030ed33</td>\n",
       "      <td>475</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44d2abe2175df8153f465f6c39b68b76a0d40ab9</th>\n",
       "      <td>Long Short-Term Memory</td>\n",
       "      <td>1997</td>\n",
       "      <td>https://doi.org/10.1162/neco.1997.9.8.1735</td>\n",
       "      <td>https://semanticscholar.org/paper/44d2abe2175df8153f465f6c39b68b76a0d40ab9</td>\n",
       "      <td>20985</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bb1a17010254abfa5e1f2a17553582ce449f8e16</th>\n",
       "      <td>Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images</td>\n",
       "      <td>2015</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/bb1a17010254abfa5e1f2a17553582ce449f8e16</td>\n",
       "      <td>323</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>600bfe5f0597ebd84898f0c4270ddfb3750594f5</th>\n",
       "      <td>Imagination-Augmented Agents for Deep Reinforcement Learning</td>\n",
       "      <td>2017</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/600bfe5f0597ebd84898f0c4270ddfb3750594f5</td>\n",
       "      <td>197</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39b19ea254b0952f2abd23ad899420749816bb1d</th>\n",
       "      <td>The Predictron: End-To-End Learning and Planning</td>\n",
       "      <td>2017</td>\n",
       "      <td></td>\n",
       "      <td>https://semanticscholar.org/paper/39b19ea254b0952f2abd23ad899420749816bb1d</td>\n",
       "      <td>113</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                                                             title  \\\n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  Human-level control through deep reinforcement learning                                                    \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  Reinforcement Learning: An Introduction                                                                    \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  The Arcade Learning Environment: An Evaluation Platform for General Agents (Extended Abstract)             \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd  Asynchronous Methods for Deep Reinforcement Learning                                                       \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813  Action-Conditional Video Prediction using Deep Networks in Atari Games                                     \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca  Prioritized Experience Replay                                                                              \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a  PILCO: A Model-Based and Data-Efficient Approach to Policy Search                                          \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e  Deep Reinforcement Learning with Double Q-Learning                                                         \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492  Reinforcement learning for robots using neural networks                                                    \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  Mastering the game of Go with deep neural networks and tree search                                         \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  Bandit Based Monte-Carlo Planning                                                                          \n",
       "2319a491378867c7049b3da055c5df60e1671158  Playing Atari with Deep Reinforcement Learning                                                             \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  Learning to Predict by the Methods of Temporal Differences                                                 \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81  Dueling Network Architectures for Deep Reinforcement Learning                                              \n",
       "282001869bd502c7917db8b32b75593addfbbc68  Neural Fitted Q Iteration - First Experiences with a Data Efficient Neural Reinforcement Learning Method   \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33  Reinforcement Learning with Unsupervised Auxiliary Tasks                                                   \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  Long Short-Term Memory                                                                                     \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16  Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images                       \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5  Imagination-Augmented Agents for Deep Reinforcement Learning                                               \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d  The Predictron: End-To-End Learning and Planning                                                           \n",
       "\n",
       "                                          year  \\\n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  2015   \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  2005   \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  2013   \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd  2016   \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813  2015   \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca  2015   \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a  2011   \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e  2016   \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492  1992   \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  2016   \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  2006   \n",
       "2319a491378867c7049b3da055c5df60e1671158  2013   \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  2005   \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81  2016   \n",
       "282001869bd502c7917db8b32b75593addfbbc68  2005   \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33  2017   \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  1997   \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16  2015   \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5  2017   \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d  2017   \n",
       "\n",
       "                                                                             doi_url  \\\n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  https://doi.org/10.1038/nature14236          \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  https://doi.org/10.1109/TNN.1998.712192      \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  https://doi.org/10.1613/jair.3912            \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd                                               \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813                                               \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca                                               \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a                                               \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e                                               \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492                                               \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  https://doi.org/10.1038/nature16961          \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  https://doi.org/10.1007/11871842_29          \n",
       "2319a491378867c7049b3da055c5df60e1671158                                               \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  https://doi.org/10.1023/A%3A1022633531479    \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81                                               \n",
       "282001869bd502c7917db8b32b75593addfbbc68  https://doi.org/10.1007/11564096_32          \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33                                               \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  https://doi.org/10.1162/neco.1997.9.8.1735   \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16                                               \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5                                               \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d                                               \n",
       "\n",
       "                                                                                                              s2_url  \\\n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  https://semanticscholar.org/paper/e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d   \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  https://semanticscholar.org/paper/97efafdb4a3942ab3efba53ded7413199f79c054   \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  https://semanticscholar.org/paper/f82e4ff4f003581330338aaae71f60316e58dd26   \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd  https://semanticscholar.org/paper/69e76e16740ed69f4dc55361a3d319ac2f1293dd   \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813  https://semanticscholar.org/paper/e4257bc131c36504a04382290cbc27ca8bb27813   \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca  https://semanticscholar.org/paper/c6170fa90d3b2efede5a2e1660cb23e1c824f2ca   \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a  https://semanticscholar.org/paper/60b7d47758a71978e74edff6dd8dea4d9c791d7a   \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e  https://semanticscholar.org/paper/3b9732bb07dc99bde5e1f9f75251c6ea5039373e   \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492  https://semanticscholar.org/paper/54c4cf3a8168c1b70f91cf78a3dc98b671935492   \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  https://semanticscholar.org/paper/846aedd869a00c09b40f1f1f35673cb22bc87490   \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  https://semanticscholar.org/paper/e635d81a617d1239232a9c9a11a196c53dab8240   \n",
       "2319a491378867c7049b3da055c5df60e1671158  https://semanticscholar.org/paper/2319a491378867c7049b3da055c5df60e1671158   \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  https://semanticscholar.org/paper/6ce57ab17fcd507b856a79874063b59555c76b3a   \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81  https://semanticscholar.org/paper/4c05d7caa357148f0bbd61720bdd35f0bc05eb81   \n",
       "282001869bd502c7917db8b32b75593addfbbc68  https://semanticscholar.org/paper/282001869bd502c7917db8b32b75593addfbbc68   \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33  https://semanticscholar.org/paper/d7bd6e3addd8bc8e2e154048300eea15f030ed33   \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  https://semanticscholar.org/paper/44d2abe2175df8153f465f6c39b68b76a0d40ab9   \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16  https://semanticscholar.org/paper/bb1a17010254abfa5e1f2a17553582ce449f8e16   \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5  https://semanticscholar.org/paper/600bfe5f0597ebd84898f0c4270ddfb3750594f5   \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d  https://semanticscholar.org/paper/39b19ea254b0952f2abd23ad899420749816bb1d   \n",
       "\n",
       "                                          num_citations  \\\n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  6647            \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  18170           \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  1003            \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd  2365            \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813  444             \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca  864             \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a  642             \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e  1388            \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492  529             \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  4912            \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  1593            \n",
       "2319a491378867c7049b3da055c5df60e1671158  2793            \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  2365            \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81  838             \n",
       "282001869bd502c7917db8b32b75593addfbbc68  557             \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33  475             \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  20985           \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16  323             \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5  197             \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d  113             \n",
       "\n",
       "                                          subgraph_citation_count  \n",
       "e0e9a94c4a6ba219e768b4e59f72c18f0a22e23d  15                       \n",
       "97efafdb4a3942ab3efba53ded7413199f79c054  13                       \n",
       "f82e4ff4f003581330338aaae71f60316e58dd26  12                       \n",
       "69e76e16740ed69f4dc55361a3d319ac2f1293dd  11                       \n",
       "e4257bc131c36504a04382290cbc27ca8bb27813  7                        \n",
       "c6170fa90d3b2efede5a2e1660cb23e1c824f2ca  7                        \n",
       "60b7d47758a71978e74edff6dd8dea4d9c791d7a  7                        \n",
       "3b9732bb07dc99bde5e1f9f75251c6ea5039373e  7                        \n",
       "54c4cf3a8168c1b70f91cf78a3dc98b671935492  6                        \n",
       "846aedd869a00c09b40f1f1f35673cb22bc87490  6                        \n",
       "e635d81a617d1239232a9c9a11a196c53dab8240  6                        \n",
       "2319a491378867c7049b3da055c5df60e1671158  6                        \n",
       "6ce57ab17fcd507b856a79874063b59555c76b3a  6                        \n",
       "4c05d7caa357148f0bbd61720bdd35f0bc05eb81  5                        \n",
       "282001869bd502c7917db8b32b75593addfbbc68  5                        \n",
       "d7bd6e3addd8bc8e2e154048300eea15f030ed33  5                        \n",
       "44d2abe2175df8153f465f6c39b68b76a0d40ab9  5                        \n",
       "bb1a17010254abfa5e1f2a17553582ce449f8e16  5                        \n",
       "600bfe5f0597ebd84898f0c4270ddfb3750594f5  5                        \n",
       "39b19ea254b0952f2abd23ad899420749816bb1d  5                        "
      ]
     },
     "execution_count": 228,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from collections import Counter\n",
    "\n",
    "c = Counter()\n",
    "process_recursively(root, lambda p: c.update([p[\"id\"]]))\n",
    "paper_counts_df = pd.DataFrame.from_records(c.most_common(), columns=[\"id\", \"subgraph_citation_count\"], index=\"id\")\n",
    "paper_counts_df = pd.concat([papers_df, paper_counts_df], axis=1)\n",
    "paper_counts_df.sort_values(\"subgraph_citation_count\", ascending=False).head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.8.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}