Neural Networks with Keras Cookbook

Validating the CNN output

Now that we have fit the model, let's validate the output we obtain from the model by implementing the feedforward portion of the CNN:

1. Let's extract the order in which weights and biases are presented:

model.weights

You can see that the weights of the convolution layer are presented first, then the bias, and finally the weight and bias in the output layer.

Also note that the shape of weights in the convolution layer is (3, 3, 1, 1) as the filter is 3 x 3 x 1 in shape (because the image is three-dimensional: 28 x 28 x 1 in shape) and the final 1 (the fourth value in shape) is for the number of filters that are specified ...