Introduction to Artificial Neural Networks
This article aims to give a beginner friendly introduction to artificial neural networks.
Firstly, it is important to note that artificial neural networks are machine learning techniques which mimic the inner workings of the human brain where interconnected neurons process information in parallel.
A neural network can be illustrated by the diagram below where each layer consists of a set of nodes, which are used to pass along signals from one layer to the next.
The inner workings of each node are designed to assign significance to the input with regards to the output that an algorithm is trying to learn. The computational process in each node involves calculating a weighted sum of input variables and passing this signal to an activation function to determine whether this signal is passed along to the next layer, and to what extent the signal will affect the final output.
The structure of a node can be illustrated using the diagram below;
To illustrate the working of a neural network, we will explore Feed Forward Neural Networks (FFNNs) in the next section.
FFNNs are the simplest forms of neural networks where nodes are only connected across layers, and nodes in the same layer do not share information, therefore, information is only passed along forward. They are simple to maintain and desirable when the input data is noisy.
Radial Basis Function Networks (RBFNs) are a form of FFNNs which perform classification by measuring the similarity of an input to a prototype from the training set in each neuron. See illustration below;
The output layer consists of a node for each category we are trying to classify, and each category is determined based on the score obtained from calculating the weighted sum of the outputs from the neurons.
Several other types of neural networks have emerged over time such as Multilayer Perceptrons where each node is connected to every node in the next layer, Recurrent Neural Networks where outputs are fed back to inputs, Convolutional Neural Networks which apply filters between layers among many others. For more information on the different types of neural networks, checkout this link.
I hope this article gives a helpful introduction to the basics of neural networks. Happy Learning :-)