3 Convolutional neural networks

Previously, we talked about artificial neural networks (ANNs), also known as multilayer perceptrons (MLPs), which are basically layers of neurons stacked on top of each other that have learnable weights and biases. Each neuron receives some inputs, which are multiplied by their weights, with nonlinearity applied via activation functions. In this chapter, we will talk about convolutional neural networks (CNNs), which are considered an evolution of the MLP architecture that performs a lot better with images.

The high-level layout of this chapter is as follows:

3.1 Image classification using MLP

3.1.1 Input layer

3.1.2 Hidden layers

3.1.3 Output layer

3.1.4 Putting it all together

3.1.5 Drawbacks of MLPs for processing images

3.2 CNN architecture

3.2.1 The big picture

3.2.2 A closer look at feature extraction

3.2.3 A closer look at classification

3.3 Basic components of a CNN

3.3.1 Convolutional layers

3.3.2 Pooling layers or subsampling

3.3.3 Fully connected layers

3.4 Image classification using CNNs

3.4.1 Building the model architecture

3.4.2 Number of parameters (weights)

3.5 Adding dropout layers to avoid overfitting

3.5.1 What is overfitting?

3.5.2 What is a dropout layer?

3.5.3 Why do we need dropout layers?