notebooks/04 Custom Estimators.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Custom Estimators\n",
    "\n",
    "## https://scikit-learn.org/dev/developers/develop.html"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Function Transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import FunctionTransformer\n",
    "ft = FunctionTransformer(func=lambda X: X+2, inverse_func=lambda X: X-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "X = np.arange(100).reshape(25, 4)\n",
    "X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ft.fit(X)\n",
    "ft.transform(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ft.inverse_transform(ft.transform(X))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Basic Transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "from sklearn.utils.validation import check_array\n",
    "\n",
    "class MyTransformer(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, first_parameter=1, second_parameter=2):\n",
    "        # all parameters must be specified in the __init__ function\n",
    "        # and init is not allowed to do anythin\n",
    "        self.first_parameter = first_parameter\n",
    "        self.second_parameter = second_parameter\n",
    "        \n",
    "    def fit(self, X, y=None):\n",
    "        X = check_array(X)\n",
    "        # fit should only take X and y as parameters\n",
    "        # even if your model is unsupervised, you need to accept a y argument!\n",
    "        \n",
    "        # Model fitting code goes here\n",
    "        print(\"fitting the model right here\")\n",
    "        # fit returns self\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X):\n",
    "        X = check_array(X)\n",
    "\n",
    "        # transform takes as parameter only X\n",
    "        \n",
    "        # apply some transformation to X:\n",
    "        X_transformed = X + 1\n",
    "        return X_transformed\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.utils.estimator_checks import check_estimator\n",
    "check_estimator(MyTransformer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import TransformerMixin\n",
    "from sklearn.utils.validation import check_X_y\n",
    "\n",
    "class MyTransformer(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, my_parameter=\"stuff\"):\n",
    "        self.my_parameter = my_parameter\n",
    "        \n",
    "    def fit(self, X, y):\n",
    "        X, y = check_X_y(X, y)\n",
    "        self.n_features_ = X.shape[1]\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X):\n",
    "        X = check_array(X)\n",
    "        if X.shape[1] != self.n_features_:\n",
    "            raise ValueError(\"Wrong number of features {} != {}\".format(\n",
    "                X.shape[1], self.n_features_))\n",
    "        return X - 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "check_estimator(MyTransformer)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exercise\n",
    "- Reimplement a simple version of the standard scaler (that removes mean and scales to unit variance) with scikit-learn interface. Can you make it pass the tests? Does it give the same result as sklearn.preprocessing.StandardScaler?\n",
    "- Reimplement a one nearest neighbor classifier with scikit-learn interface (that memorizes the training set and assignes a new test point to the class of the closest training point). Again, try making it pass the tests.\n",
    "\n",
    "hint: use sklearn.utils.validation.check_is_fitted and sklearn.utils.validation.unique_labels (though you don't have to)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load solutions/custom_estimators.py"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}