Before training an artificial neural network, we must first understand what an artificial neural network is. For this reason, I have implemented a modular multilayer perceptron model that will help us understand how a neural network works and how we can use it to diagnose breast cancer.
A perceptron is an artificial neuron. They are defined by the presence of one or more input connections, an activation function and a single output. Each connection contains a weight which is learned during the training phase.
Its main function is to replicate the behavior of a biological neuron. Broadly speaking, we can see that a neuron can be divided into four parts:
- Dendrites: where a neuron receives input from other cells.
- Synapse: points of contact between neurons where information is passed from one neuron to the next.
- Soma: where the signals from the dendrites are joined and passed on.
- Axons: where electrical impulses from the neuron travel away to be received by other neurons.
Now that we understand how a biological neuron works, we can find its similarities to an artificial neuron:
- Inputs ~ Dendrites
- Weights ~ Synapse
- Soma ~ Node
- Output ~ Axon
The multilayer perceptron is a feedforward network. This means that the data flows from the input layer to the output layer. A multilayer perceptron is defined by the presence of one or more hidden layers as well as an interconnection of all the neurons of one layer to the next.
