{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from scipy.sparse.linalg import svds\n",
    "from sklearn.datasets import load_iris\n",
    "from sklearn.decomposition import TruncatedSVD as skTruncatedSVD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class TruncatedSVD():\n",
    "    def __init__(self, n_components=2):\n",
    "        self.n_components = n_components\n",
    "\n",
    "    def fit(self, X):\n",
    "        U, Sigma, VT = svds(X, k=self.n_components)\n",
    "        U, Sigma, VT = U[:, ::-1], Sigma[::-1], VT[::-1]\n",
    "        self.components_ = VT\n",
    "        self.singular_values_ = Sigma\n",
    "        self.explained_variance_ = np.var(U * Sigma, axis=0)\n",
    "        self.explained_variance_ratio_ = self.explained_variance_ / np.var(X, axis=0).sum()\n",
    "        return self\n",
    "\n",
    "    def transform(self, X):\n",
    "        return np.dot(X, self.components_.T)\n",
    "\n",
    "    def inverse_transform(self, X):\n",
    "        return np.dot(X, self.components_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, _ = load_iris(return_X_y=True)\n",
    "svd1 = TruncatedSVD().fit(X)\n",
    "Xt1 = svd1.transform(X)\n",
    "Xinv1 = svd1.inverse_transform(Xt1)\n",
    "svd2 = skTruncatedSVD(n_components=2).fit(X)\n",
    "Xt2 = svd2.transform(X)\n",
    "Xinv2 = svd2.inverse_transform(Xt2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(svd1.components_.shape[0]):\n",
    "    assert np.allclose(svd1.components_[i], svd2.components_[i]) or np.allclose(svd1.components_[i], -svd2.components_[i])\n",
    "assert np.allclose(svd1.singular_values_, svd2.singular_values_)\n",
    "assert np.allclose(svd1.explained_variance_, svd2.explained_variance_)\n",
    "assert np.allclose(svd1.explained_variance_ratio_, svd2.explained_variance_ratio_)\n",
    "for i in range(Xt1.shape[1]):\n",
    "    assert np.allclose(Xt1[:, i], Xt2[:, i]) or np.allclose(Xt1[:, i], -Xt2[:, i])\n",
    "assert np.allclose(Xinv1, Xinv2)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "dev",
   "language": "python",
   "name": "dev"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}