Preparing models and data

To produce explanations, nnitp needs both the model and the data from which the model was trained. To provide this information to nnitp, prepare a Python file with a name like foo_model.py, where foo is the model name. This file should contain these items:

• function get_data() returns the training and test datasets in this form:

  (x_train, y_train), (x_test,y_test)
  

  where

  • x_train is the vector of training input samples
  • y_train is the vector of training sample labels
  • x_test is the vector of test input samples
  • y_test is the vector of test sample labels.

  All of these should be represented as numpy arrays.

  The input data should be the exact data used to train the model. For example, if some preprocessing steps were applied to the original dataset, the same preprocessing should be applied by get_data.

• function get_model() returns the model, in an appropriate wrapper

  The wrapper used depends on the model type:

  • For Keras models, use nnitp.keras.Wrapper.

• dictionary params mapping parameter names to values.

As an example, here is a Python file mnist_model.py that loads a Keras model of the MNIST digit recognition dataset from a file mnist_model.h5:

import tensorflow as tf
from keras.models import load_model
from nnitp.keras import Wrapper

# Fetch the MNIST dataset, normalize to range [0.0,1.0)

def get_data():
    (x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
    x_train = x_train.reshape(x_train.shape[0], 28, 28, 1)
    x_test = x_test.reshape(x_test.shape[0], 28, 28, 1)
    x_train = x_train.astype('float32')
    x_test = x_test.astype('float32')
    x_train /= 255
    x_test /= 255
    return (x_train, y_train), (x_test, y_test)

# Fetch the trained model

def get_model():
    return Wrapper(load_model('mnist_model.h5'))

params = {'size':100,'alpha':0.98,'gamma':0.6,'mu':0.9,'layers':[2]}

This file should be found in the nnitp’s load path, which includes the current working directory and source subdirectory models.

Models found in this path can be selected in the GUI.