{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from scipy.sparse.linalg import lsqr\n",
    "from sklearn.datasets import load_iris, load_breast_cancer\n",
    "from sklearn.linear_model import RidgeClassifier as skRidgeClassifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Implementation 1\n",
    "- convert classification problem to regression problem through binarizing labels in a one-vs-all fashion\n",
    "- based on scipy.sparse.linalg.lsqr (see the implementation of linear_model.Ridge)\n",
    "- center the dataset and calculate the intercept manually\n",
    "- similar to scikit-learn solver='lsqr'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class RidgeClassifier:\n",
    "    def __init__(self, alpha=1.0):\n",
    "        self.alpha = alpha\n",
    "\n",
    "    def _encode(self, y):\n",
    "        classes = np.unique(y)\n",
    "        y_train = np.full((y.shape[0], len(classes)), -1)\n",
    "        for i, c in enumerate(classes):\n",
    "            y_train[y == c, i] = 1\n",
    "        if len(classes) == 2:\n",
    "            y_train = y_train[:, 1].reshape(-1, 1)\n",
    "        return classes, y_train\n",
    "\n",
    "    def _solve_lsqr(self, X, y):\n",
    "        coefs = np.zeros((y.shape[1], X.shape[1]))\n",
    "        for i in range(y.shape[1]):\n",
    "            cur_y = y[:, i]\n",
    "            info = lsqr(X, cur_y, np.sqrt(self.alpha))\n",
    "            coefs[i] = info[0]\n",
    "        return coefs\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        self.classes_, y_train = self._encode(y)\n",
    "        X_mean = np.mean(X, axis=0)\n",
    "        y_mean = np.mean(y_train, axis=0)\n",
    "        X_train = X - X_mean\n",
    "        y_train = y_train - y_mean\n",
    "        self.coef_ = self._solve_lsqr(X_train, y_train)\n",
    "        self.intercept_ = y_mean - np.dot(X_mean, self.coef_.T)\n",
    "        return self\n",
    "\n",
    "    def decision_function(self, X):\n",
    "        scores = np.dot(X, self.coef_.T) + self.intercept_\n",
    "        if scores.shape[1] == 1:\n",
    "            return scores.ravel()\n",
    "        else:\n",
    "            return scores\n",
    "\n",
    "    def predict(self, X):\n",
    "        scores = self.decision_function(X)\n",
    "        if len(scores.shape) == 1:\n",
    "            indices = (scores > 0).astype(int)\n",
    "        else:\n",
    "            indices = np.argmax(scores, axis=1)\n",
    "        return self.classes_[indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# binary classification\n",
    "for alpha in [0.1, 1, 10]:\n",
    "    X, y = load_breast_cancer(return_X_y = True)\n",
    "    clf1 = RidgeClassifier(alpha=alpha).fit(X, y)\n",
    "    clf2 = skRidgeClassifier(alpha=alpha, solver='lsqr',\n",
    "                             # keep consisent with scipy default\n",
    "                             tol=1e-8).fit(X, y)\n",
    "    assert clf1.coef_.shape == (1, X.shape[1])\n",
    "    assert np.allclose(clf1.coef_, clf2.coef_)\n",
    "    assert np.allclose(clf1.intercept_, clf2.intercept_)\n",
    "    prob1 = clf1.decision_function(X)\n",
    "    prob2 = clf2.decision_function(X)\n",
    "    pred1 = clf1.predict(X)\n",
    "    pred2 = clf2.predict(X)\n",
    "    assert np.allclose(prob1, prob2)\n",
    "    assert np.array_equal(pred1, pred2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# multiclass classification\n",
    "for alpha in [0.1, 1, 10]:\n",
    "    X, y = load_iris(return_X_y = True)\n",
    "    clf1 = RidgeClassifier(alpha=alpha).fit(X, y)\n",
    "    clf2 = skRidgeClassifier(alpha=alpha, solver='lsqr',\n",
    "                             # keep consisent with scipy default\n",
    "                             tol=1e-8).fit(X, y)\n",
    "    assert clf1.coef_.shape == (len(np.unique(y)), X.shape[1])\n",
    "    assert np.allclose(clf1.coef_, clf2.coef_)\n",
    "    assert np.allclose(clf1.intercept_, clf2.intercept_)\n",
    "    prob1 = clf1.decision_function(X)\n",
    "    prob2 = clf2.decision_function(X)\n",
    "    pred1 = clf1.predict(X)\n",
    "    pred2 = clf2.predict(X)\n",
    "    assert np.allclose(prob1, prob2)\n",
    "    assert np.array_equal(pred1, pred2)"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}