Kernel Ridge, SVR, PCA.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# load the data\n",
    "\n",
    "K = np.loadtxt('ncicancer_kernel_hf_ex0.txt')\n",
    "y = np.loadtxt('y1.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "K.shape, y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.cross_validation import ShuffleSplit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cv = ShuffleSplit(K.shape[0], n_iter=5, test_size=500, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.kernel_ridge import KernelRidge\n",
    "from sklearn.svm import SVR\n",
    "from sklearn.metrics.regression import r2_score\n",
    "\n",
    "models = KernelRidge, SVR\n",
    "\n",
    "scores = {model.__name__: [] for model in models}\n",
    "\n",
    "for outer_train, outer_test in cv:\n",
    "    for model in models:\n",
    "        est = model(kernel='precomputed')\n",
    "        K_train_train = K[outer_train[:, np.newaxis], outer_train]\n",
    "        y_train = y[outer_train]\n",
    "        \n",
    "        est.fit(K_train_train, y_train)\n",
    "        K_test_train = K[outer_test[:, np.newaxis], outer_train]\n",
    "        y_test = y[outer_test]\n",
    "        pred = est.predict(K_test_train)\n",
    "        score = r2_score(y_test, pred)\n",
    "        scores[model.__name__].append(score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "{key: (np.mean(value), np.std(value)) for key, value in scores.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "alphas = np.logspace(-3, 3, 20)\n",
    "Cs = np.logspace(3, -3, 20)\n",
    "\n",
    "models = (KernelRidge, 'alpha', alphas), (SVR, 'C', Cs)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.cross_validation import KFold\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "scores = {model[0].__name__: [] for model in models}\n",
    "\n",
    "for outer_train, outer_test in cv:\n",
    "    inner_cv = KFold(len(outer_train), n_folds=3)\n",
    "    for model, param, param_vals in models:\n",
    "        print(\"evaluating {}\".format(model.__name__))\n",
    "        inner_scores = []\n",
    "        K_train_train = K[outer_train[:, np.newaxis], outer_train]\n",
    "        y_train = y[outer_train]\n",
    "        K_test_train = K[outer_test[:, np.newaxis], outer_train]\n",
    "        y_test = y[outer_test]\n",
    "        for param_val in param_vals:\n",
    "            #print(\"testing param {}\".format(param_val))\n",
    "            param_scores = []\n",
    "            est = model(kernel='precomputed')\n",
    "            setattr(est, param, param_val)\n",
    "            for inner_train, inner_test in inner_cv:\n",
    "                est.fit(K_train_train[inner_train[:, np.newaxis],\n",
    "                                      inner_train],\n",
    "                        y_train[inner_train])\n",
    "                pred = est.predict(K_train_train[\n",
    "                            inner_test[:, np.newaxis],\n",
    "                            inner_train])\n",
    "                score = r2_score(y_train[inner_test], pred)\n",
    "                param_scores.append(score)\n",
    "            inner_scores.append(param_scores)\n",
    "        mean_inner_scores = np.mean(inner_scores, axis=1)\n",
    "        argbest_param = np.argmax(mean_inner_scores)\n",
    "        best_param = param_vals[argbest_param]\n",
    "        print(\"Best parameter for {}: {}\".format(model.__name__, best_param))\n",
    "        setattr(est, param, best_param)\n",
    "        est.fit(K_train_train, y_train)\n",
    "        pred = est.predict(K_test_train)\n",
    "        score = r2_score(y_test, pred)\n",
    "        scores[model.__name__].append(score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "{key: np.mean(value) for key, value in scores.items()}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Kernel PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_digits\n",
    "\n",
    "digits = load_digits()\n",
    "\n",
    "X, y = digits.data, digits.target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.decomposition import KernelPCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PCA(copy=True, n_components=None, whiten=False)"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pca = PCA()\n",
    "pca.fit(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWYAAADRCAYAAAD/nhhvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGYoZAJyhmAHAGYoZAJyhmAHAmf8AmIBuIdeL+7cAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa576ec2450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure()\n",
    "for i in range(pca.n_components):\n",
    "    plt.subplot(2, pca.n_components / 2, i + 1)\n",
    "    plt.imshow(pca.components_[i].reshape(8, 8), interpolation='nearest')\n",
    "    plt.gray()\n",
    "    plt.axis('off')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KernelPCA(alpha=1.0, coef0=1, degree=3, eigen_solver='auto',\n",
       "     fit_inverse_transform=True, gamma=None, kernel='linear',\n",
       "     kernel_params=None, max_iter=None, n_components=None,\n",
       "     remove_zero_eig=False, tol=0)"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kpca = KernelPCA(kernel='linear', fit_inverse_transform=True)\n",
    "kpca.fit(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEDCAYAAAA849PJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nof4/o7YuCo8SS74AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa577cdcf90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(kpca.lambdas_[:64])\n",
    "plt.plot(pca.explained_variance_)\n",
    "plt.yscale('log')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "kernels = ('rbf', 'gamma', [.01, 1.]), ('poly', 'degree', [1, 3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7fa5773f3210>"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD7CAYAAABjVUMJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngIhIgVICEBEpUP8frtd4nBJYvJQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa577846c10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "for kernel, param_name, param_vals in kernels:\n",
    "    kpca = KernelPCA(kernel=kernel)\n",
    "    for param_val in param_vals:\n",
    "        setattr(kpca, param_name, param_val)\n",
    "        kpca.fit(X)\n",
    "        plt.plot(kpca.lambdas_[:64], label='{} {} {}'.format(\n",
    "                    kernel, param_name, param_val))\n",
    "        plt.yscale('log')\n",
    "plt.legend(loc='lower left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "KernelPCA??"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## transforming and inverse transforming"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "kpca = KernelPCA(alpha=.1, kernel='rbf', gamma=.1, fit_inverse_transform=True, n_components=60)\n",
    "cv = ShuffleSplit(len(X), n_iter=1, test_size=500)\n",
    "\n",
    "train, test = next(iter(cv))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KernelPCA(alpha=0.1, coef0=1, degree=3, eigen_solver='auto',\n",
       "     fit_inverse_transform=True, gamma=0.1, kernel='rbf',\n",
       "     kernel_params=None, max_iter=None, n_components=60,\n",
       "     remove_zero_eig=False, tol=0)"
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kpca.fit(X[train])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Xtrain_pca = kpca.transform(X[train])\n",
    "Xinv = kpca.inverse_transform(Xtrain_pca)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1297, 64)"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xinv.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1297, 4)"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xtrain_pca.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/me232320/software/python_pkg/anaconda/lib/python2.7/site-packages/sklearn/metrics/regression.py:468: DeprecationWarning: Default 'multioutput' behavior now corresponds to 'variance_weighted' value, it will be changed to 'uniform_average' in 0.18.\n",
      "  DeprecationWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.085798974213989382"
      ]
     },
     "execution_count": 148,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r2_score(X[train], Xinv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1297, 64)"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xinv.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1296"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "36 ** 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fa577044d90>"
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP0AAAD7CAYAAAChbJLhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa5772ea390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.matshow(Xinv[:100].reshape(10, 10, 8, 8).transpose(0, 2, 1, 3).reshape(10 * 8, -1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}