Understanding Machine Learning (ML) Model Interpretability

Model interpretability refers to the degree of ease with which a human can comprehend the logic behind the ML model’s predictions. Essentially, it is the capability to comprehend how a model arrives at its decisions and which variables are contributing to its forecasts.

Building a CNN for image classification in Keras

Let’s see an example of model interpretability using a deep learning convolutional neural network (CNN) for image classification. We’ll build our model in Python using Keras. For this purpose, we’ll download the CIFAR-10 dataset directly from keras.datasets: it consists of 60,000 32x32 color images (so 3-channels images) in 10 classes (airplane, automobile, bird, cat, deer, dog, frog, horse, ship, and truck), with 6,000 images per class. Here, we will share just the body of the model.