examples/mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Explaining MNIST Image Classifiers with CXPlain\n",
    "\n",
    "First, we load 500 sample images from the MNIST dataset which we will use to attempt to recognise the digit type\n",
    "that a given image shows."
   ]
  },
  {
   "cell_type": "code",
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "source": [
    "import os\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'\n",
    "from cxplain.util.test_util import TestUtil\n",
    "\n",
    "num_subsamples = 500\n",
    "(x_train, y_train), (x_test, y_test) = TestUtil.get_mnist(flattened=False, num_subsamples=num_subsamples)"
   ],
   "execution_count": 1,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Next, we fit a multilayer perceptron (MLP) that predicts the digit type that a given MNIST image shows."
   ]
  },
  {
   "cell_type": "code",
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "source": [
    "from sklearn.neural_network import MLPClassifier\n",
    "\n",
    "explained_model = MLPClassifier(solver='lbfgs', alpha=1e-5,\n",
    "                                hidden_layer_sizes=(64, 32), random_state=1)\n",
    "explained_model.fit(x_train.reshape((len(x_train), -1)), y_train);"
   ],
   "execution_count": 2,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "After fitting the multilayer perceptron, we wish to explain its decisions, i.e. what input features were most relevant\n",
    "for a given model prediction. To do so, we train a causal explanation (CXPlain) model that can learn to explain any\n",
    "machine-learning model using the same training data. In practice, we have to define:\n",
    "- `model_builder`: The type of model we want to use as our CXPlain model. In this case we are using a neural explanation model using\n",
    "a UNet structure. \n",
    "- `masking_operation`: The masking operaion used to remove a certain input feature from the set of available input features. In this case we are using zero masking, i.e. setting the value of that input feature to zero.\n",
    "- `loss`: The loss function that we wish to use to measure the impact of removing a certain input feature from the set of available features. In most common use cases, this will be the mean squared error (MSE) for regression problems and the cross-entropy for classification problems.\n"
   ]
  },
  {
   "cell_type": "code",
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "source": [
    "from tensorflow.python.keras.losses import categorical_crossentropy\n",
    "from cxplain import UNetModelBuilder, ZeroMasking, CXPlain\n",
    "\n",
    "downsample_factors = (2, 2)\n",
    "model_builder = UNetModelBuilder(downsample_factors, num_layers=2, num_units=8, activation=\"relu\",\n",
    "                                 p_dropout=0.0, verbose=0, batch_size=32, learning_rate=0.001)\n",
    "masking_operation = ZeroMasking()\n",
    "loss = categorical_crossentropy"
   ],
   "execution_count": 3,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Using this configuration, we now instantiate a CXPlain model and fit it to the same Boston Housing data that we used to fit the MLP model that we wish to explain.\n",
    "We additionally request that 5 bootstrap resampled versions of this model are trained in order to be able to compute confidence intervals for the estimated importance scores.\n",
    "Only a single model is trained - and computation of confidence intervals is not possible - if you do not specify the number of models you wish to train."
   ]
  },
  {
   "cell_type": "code",
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "source": [
    "explainer = CXPlain(explained_model, model_builder, masking_operation, loss, \n",
    "                    num_models=5, downsample_factors=downsample_factors, flatten_for_explained_model=True)\n",
    "explainer.fit(x_train, y_train);"
   ],
   "execution_count": 4,
   "outputs": [
    {
     "name": "stdout",
     "text": [
      "WARNING:tensorflow:From /Users/schwabp3/Documents/projects/venv/lib/python2.7/site-packages/tensorflow/python/ops/init_ops.py:1251: calling __init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\nInstructions for updating:\nCall initializer instance with the dtype argument instead of passing it to the constructor\n",
      "WARNING:tensorflow:From /Users/schwabp3/Documents/projects/cxplain-github/cxplain/backend/model_builders/base_model_builder.py:154: The name tf.train.AdamOptimizer is deprecated. Please use tf.compat.v1.train.AdamOptimizer instead.\n\n",
      "WARNING:tensorflow:From /Users/schwabp3/Documents/projects/venv/lib/python2.7/site-packages/tensorflow/python/ops/math_grad.py:1250: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nUse tf.where in 2.0, which has the same broadcast rule as np.where\n"
     ],
     "output_type": "stream"
    }
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "We can then use this fitted CXPlain model to explain the predictions of the explained model on the held-out test samples.\n",
    "We additionally request confidence intervals at the 80% confidence level, i.e. the 10% and 90% quantiles of the observed output distribution of importance scores.\n",
    "Note that the importance scores are normalised to sum to a value of 1 and each score therefore represents the relative importance of each respective input feature."
   ]
  },
  {
   "cell_type": "code",
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "source": [
    "attributions, confidence = explainer.explain(x_test, confidence_level=0.80)"
   ],
   "execution_count": 5,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "We can now visualise the 2x2 pixel block attributions and their confidence for a specific sample from the test set using the `Plot` toolset available as part of CXPlain."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": "<Figure size 640x480 with 3 Axes>"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 0 Axes>"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from cxplain.visualisation.plot import Plot\n",
    "\n",
    "plt.rcdefaults()\n",
    "\n",
    "selected_index = 2\n",
    "selected_sample = x_test[selected_index]\n",
    "importances = attributions[selected_index]\n",
    "importances_confidence = confidence[selected_index]\n",
    "\n",
    "Plot.plot_attribution_2d(selected_sample,\n",
    "                         importances, \n",
    "                         importances_confidence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": "<Figure size 640x480 with 3 Axes>"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 0 Axes>"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "selected_index = 3\n",
    "selected_sample = x_test[selected_index]\n",
    "importances = attributions[selected_index]\n",
    "importances_confidence = confidence[selected_index]\n",
    "\n",
    "Plot.plot_attribution_2d(selected_sample,\n",
    "                         importances, \n",
    "                         importances_confidence)\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n",
     "is_executing": false
    }
   }
  }
 ],
 "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": "2.7.6"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "source": [],
    "metadata": {
     "collapsed": false
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}