ABSTRACT
The working of biological neurons inspires artificial neural networks (ANNs), another widely used supervised learning algorithm. An ANN consists of three layers: the input layer, the hidden layer, and the output layer. A neuron produces output by calculating the weighted sum of the input features, which then passes through an activation function. The activation function works similar to a gate to control the flow of information from one layer to another. Neurons learn by updating their weights during training. Weights are adjusted by using an algorithm known as gradient descent. The gradient descent algorithm uses derivatives to backpropagate the errors and updates the weights in such a manner that the performance of the network improves. The working of ANNs will be discussed in detail in this chapter, and further Python implementation of ANNs will be shown using Tensorflow’s Keras library. The chapter will also discuss the applications of convoluted neural networks (CNNs) on image data. While ANNs learn features individually, CNNs can learn local spatial features. A CNN-based model is trained for identifying plasmodium-infected blood smears obtained from light microscope image data.
