4_2_spacenet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Crash course on Nilearn and MVPA (Multi-Variate Pattern Analysis)\n",
    "In this lesson, we will cover some popular methods in MVPA in the context of brain imaging. Namely,\n",
    "* <a href=\"http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1383651/\">Searchlight [Nikolaus et al. 2006 (PNAS)]</a>\n",
    "* <a href=\"https://hal.inria.fr/hal-01147731/document\">Space-Net [Dohmatob et al. 2014 (PRNI)]</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load Haxby 2001 data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from nilearn.datasets import fetch_haxby\n",
    "data_files = fetch_haxby()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Load Target labels\n",
    "import numpy as np\n",
    "labels = np.recfromcsv(data_files.session_target[0], delimiter=\" \")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "rec.array([('rest', 0), ('rest', 0), ('rest', 0), ..., ('rest', 11), ('rest', 11),\n",
       " ('rest', 11)], \n",
       "          dtype=[('labels', 'S12'), ('chunks', '<i8')])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Restrict to face and house conditions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "target = labels['labels']\n",
    "condition_mask = np.logical_or(target == b\"face\", target == b\"house\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "Split data into train and test pools"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "condition_mask_train = np.logical_and(condition_mask, labels['chunks'] <= 6)\n",
    "condition_mask_test = np.logical_and(condition_mask, labels['chunks'] > 6)\n",
    "from nilearn.image import index_img\n",
    "from nilearn.masking import compute_epi_mask\n",
    "func_filenames = data_files.func[0]\n",
    "X_train = index_img(func_filenames, condition_mask_train)\n",
    "X_test = index_img(func_filenames, condition_mask_test)\n",
    "y_train = target[condition_mask_train]\n",
    "y_test = target[condition_mask_test]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Compute brain mask and background EPI image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from nilearn.image import mean_img\n",
    "mask_img = compute_epi_mask(func_filenames)\n",
    "background_img = mean_img(func_filenames)\n",
    "background_img.to_filename(\"bg_img.nii.gz\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Searchlight\n",
    "============"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prepare masks\n",
    "\n",
    "* mask_img is the original mask\n",
    "* process_mask_img is a subset of mask_img, it contains the voxels that should be processed (we only keep the slice z = 26 and the back of the brain to speed up computation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from nilearn.decoding import SearchLight\n",
    "from nilearn.image import new_img_like\n",
    "process_mask = mask_img.get_data().astype(np.int)\n",
    "picked_slice = 27\n",
    "process_mask[..., (picked_slice + 1):] = 0\n",
    "process_mask[..., :picked_slice] = 0\n",
    "process_mask[:, 30:] = 0\n",
    "process_mask_img = new_img_like(mask_img, process_mask)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, fit a Searchlight model on your data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    8.4s finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "SearchLight(cv=None, estimator='svc',\n",
       "      mask_img=<nibabel.nifti1.Nifti1Image object at 0x7f809d75af10>,\n",
       "      n_jobs=1,\n",
       "      process_mask_img=<nibabel.nifti1.Nifti1Image object at 0x7f8070e40f90>,\n",
       "      radius=5.6, scoring=None, verbose=1)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "searchlight = SearchLight(mask_img, process_mask_img=process_mask_img, radius=5.6, verbose=1)\n",
    "searchlight.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot the model coefficients to see whats really going on."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<nilearn.plotting.displays.YZSlicer at 0x7f806a69a350>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaMAAADXCAYAAAC3UT6XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f806a755c10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from nilearn.plotting import plot_stat_map, show\n",
    "coef_img = new_img_like(background_img, searchlight.scores_)\n",
    "plot_stat_map(coef_img, background_img,\n",
    "              title=\"Search Light\",\n",
    "              cut_coords=(-34, -16), display_mode=\"yz\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SpaceNet\n",
    "========\n",
    "* This is a family of priors which simultaneously impose sparsity and structure on the maps of regressor coefficients.\n",
    "* It includes <a href=\"http://www.ncbi.nlm.nih.gov/pubmed/23298747\">GraphNet [Grosenick et al 2013]</a>, <a href=\"https://hal.inria.fr/hal-00839984\">TV-L1 [Gramfort et al. 2013 (PRNI)]</a>, etc.\n",
    "\n",
    "![caption](haxby_comp.png)\n",
    "\n",
    "\n",
    "Formally, the model is\n",
    "\\begin{equation}\n",
    "\\text{minimize}_{\\bf{w} \\in \\mathbb R^p}\\mathcal L(y, X\\bf{w}) + \\alpha (\\rho \\|\\bf{w}\\|_1 + (1-\\rho) \\Omega(\\bf{w}))\n",
    "\\end{equation}\n",
    "where\n",
    "* $\\bf{w} \\in \\mathbb R^p$ is the vector of model coefficients and $p$ is the number of voxels in the brain mask (say, $p \\sim 10^6$ in a full brain mask!)\n",
    "* $\\mathcal L(y, X\\bf{w})$ is the loss term. For example $\\mathcal L(y, X\\bf{w}) = \\frac{1}{2}\\|X\\bf{w} - y\\|_2^2$ in regression settings, equals the logistic loss in classification settings, etc.\n",
    "* $\\alpha > 0$ is the regularization parameter and controls and controls the relative importance of the penalty and loss terms. $\\rho \\in [0, 1]$ is a mixing parameter and controls the tradeoff between sparsity and structure. For example  $\\rho = 1$ corresponds to the classical LASSO.\n",
    "* The penalty term $\\|\\bf{w}\\|_1$ induces sparsity on the model coefficients $\\bf{w}$: all coefficients are forced to zero, except the most \"predictive\" ones!\n",
    "* $\\Omega(\\bf{w})$ is a structure inducing penalty on the model coefficients:\n",
    "\n",
    "For example \\begin{equation}\\Omega(\\bf{w}) = \\frac{1}{2}\\|\\nabla \\bf{w}\\|_F^2 := \\sum_{j=1}^p|(\\nabla_x \\bf{w})_j|^2 + |(\\nabla_y \\bf{w})_j|^2 + |(\\nabla_z \\bf{w})_j|^2\\end{equation} in GraphNet / S-Lasso and for TV-L1, we have\n",
    "\\begin{equation}\\Omega(\\bf{w}) = \\|\\bf{w}\\|_{TV} := \\sum_{j=1}^p\\sqrt{|(\\nabla_x \\bf{w})_j|^2 + |(\\nabla_y \\bf{w})_j|^2 + |(\\nabla_z \\bf{w})_j|^2}\\end{equation}\n",
    "\n",
    "* GraphNet is Tikhonov regularization on the spatial gradient of the model coefficients $\\bf{w}$. It will produce smooth blobs, but no sharp edges.\n",
    "* TV semi-norm promotes smooth blobs with sharp edges (i.e has a segmentation property, which is a very good thing). However, the resulting model is much harder to optimize than GraphNet."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**GraphNet / S-Lasso**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from nilearn.decoding import SpaceNetClassifier\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  8, ..., 125]), array([ 0,  1,  2,  3,  4,  5,  6,  7,  9, 10, 11, 12, 13, 14, 15, 16]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 0), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "......."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 4.7s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 8, 17, 18, 19, 20, 21, 22, 23, 24, 27, 28, 29, 30, 31, 32, 33]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 1), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 4.9s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([25, 26, 34, 35, 36, 37, 38, 39, 40, 41, 45, 46, 47, 48, 49, 50]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 2), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 8.5s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([42, 43, 44, 51, 52, 53, 54, 55, 56, 57, 58, 63, 64, 65, 66, 67]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 3), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 6.1s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([59, 60, 61, 62, 68, 69, 70, 71, 72, 73, 74, 75, 81, 82, 83, 84]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 4), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 5.2s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 76,  77,  78,  79,  80,  85,  86,  87,  88,  89,  90,  91,  92,\n",
      "        99, 100, 101]), \n",
      "{ 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 5), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "......."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 9.8s, 0.2min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 93,  94,  95,  96,  97,  98, 102, 103, 104, 105, 106, 107, 108,\n",
      "       109, 117, 118]), \n",
      "{ 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 6), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 8.5s, 0.1min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<function _graph_net_squared_loss at 0x7f806c7780c8>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 118]), array([110, 111, 112, 113, 114, 115, 116, 119, 120, 121, 122, 123, 124, 125]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 7), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "______________________________________________________path_scores - 4.7s, 0.1min\n",
      "Time Elapsed: 53.5496 seconds, 0 minutes.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   8 out of   8 | elapsed:   52.6s finished\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Graph-net classification accuracy : 97.7778%\n"
     ]
    }
   ],
   "source": [
    "decoder = SpaceNetClassifier(memory=\"nilearn_cache\", memory_level=2, penalty='graph-net', verbose=1, loss=\"mse\")\n",
    "decoder.fit(X_train, y_train)\n",
    "y_pred = decoder.predict(X_test)\n",
    "accuracy = (y_pred == y_test).mean() * 100.\n",
    "print(\"Graph-net classification accuracy : %g%%\" % accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<nilearn.plotting.displays.YZSlicer at 0x7f806a33add0>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaMAAADXCAYAAAC3UT6XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nckkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f806aa00690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from nilearn.plotting import plot_stat_map, show\n",
    "coef_img = decoder.coef_img_\n",
    "plot_stat_map(coef_img, background_img,\n",
    "              title=\"graph-net: accuracy %g%%\" % accuracy,\n",
    "              cut_coords=(-34, -16), display_mode=\"yz\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**TV-L1**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "decoder = SpaceNetClassifier(memory=\"nilearn_cache\", memory_level=2, penalty='tv-l1', loss=\"mse\", verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  8, ..., 125]), array([ 0,  1,  2,  3,  4,  5,  6,  7,  9, 10, 11, 12, 13, 14, 15, 16]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 0), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 15.1s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 8, 17, 18, 19, 20, 21, 22, 23, 24, 27, 28, 29, 30, 31, 32, 33]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 1), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 17.7s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([25, 26, 34, 35, 36, 37, 38, 39, 40, 41, 45, 46, 47, 48, 49, 50]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 2), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 20.0s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([42, 43, 44, 51, 52, 53, 54, 55, 56, 57, 58, 63, 64, 65, 66, 67]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 3), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "____________________________________________________path_scores - 110.4s, 1.8min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([59, 60, 61, 62, 68, 69, 70, 71, 72, 73, 74, 75, 81, 82, 83, 84]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 4), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 18.9s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 76,  77,  78,  79,  80,  85,  86,  87,  88,  89,  90,  91,  92,\n",
      "        99, 100, 101]), \n",
      "{ 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 5), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 20.4s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 125]), array([ 93,  94,  95,  96,  97,  98, 102, 103, 104, 105, 106, 107, 108,\n",
      "       109, 117, 118]), \n",
      "{ 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 6), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      ".........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 18.6s, 0.3min\n",
      "________________________________________________________________________________\n",
      "[Memory] Calling nilearn.decoding.space_net.path_scores...\n",
      "path_scores(<functools.partial object at 0x7f806a1ece10>, array([[ 1.280958, ...,  0.069205],\n",
      "       ..., \n",
      "       [-0.421694, ..., -2.454092]], dtype=float32), \n",
      "array([-1, ...,  1]), array([[[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]],\n",
      "\n",
      "       ..., \n",
      "       [[False, ..., False],\n",
      "        ..., \n",
      "        [False, ..., False]]], dtype=bool), \n",
      "None, [0.5], array([  0, ..., 118]), array([110, 111, 112, 113, 114, 115, 116, 119, 120, 121, 122, 123, 124, 125]), { 'max_iter': 1000, 'tol': 0.0001}, debias=False, verbose=1, n_alphas=10, eps=0.001, key=(0, 7), screening_percentile=30.929715190250057, is_classif=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "........."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_____________________________________________________path_scores - 20.1s, 0.3min\n",
      "Time Elapsed: 242.163 seconds, 4 minutes.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   8 out of   8 | elapsed:  4.0min finished\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n",
      "/home/elvis/anaconda2/lib/python2.7/site-packages/sklearn/externals/joblib/hashing.py:197: DeprecationWarning: Changing the shape of non-C contiguous array by\n",
      "descriptor assignment is deprecated. To maintain\n",
      "the Fortran contiguity of a multidimensional Fortran\n",
      "array, use 'a.T.view(...).T' instead\n",
      "  obj_bytes_view = obj.view(self.np.uint8)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Graph-net classification accuracy : 97.7778%\n"
     ]
    }
   ],
   "source": [
    "decoder.fit(X_train, y_train)\n",
    "y_pred = decoder.predict(X_test)\n",
    "accuracy = (y_pred == y_test).mean() * 100.\n",
    "print(\"Graph-net classification accuracy : %g%%\" % accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<nilearn.plotting.displays.YZSlicer at 0x7f806a097210>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaMAAADXCAYAAAC3UT6XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGJ5PYgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f806a3c4c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from nilearn.plotting import plot_stat_map, show\n",
    "coef_img = decoder.coef_img_\n",
    "plot_stat_map(coef_img, background_img,\n",
    "              title=\"graph-net: accuracy %g%%\" % accuracy,\n",
    "              cut_coords=(-34, -16), display_mode=\"yz\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n"
   ]
  }
 ],
 "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": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}