Artificial Neural Networks (ANN) are machine learning models that try to mimic the functions of a biological brain. A fundamental unit of a neural network is a structure called neuron. A neuron has one or multiple inputs and a soma: a part of the neuron that sums up the input signals that are then passed through an activation (transition) function that decides whether (and/or how) to propagate the signal to the output. There is a multitude of various transition functions that an artificial neuron can implement. These range from the basic ones such as the step function that sends a signal only if a certain threshold is exceeded through linear activation functions that do not alter the signal in any way to some nonlinear functions such as tanh
, sigmoid
, or RBF
.
A neural network is a structure that combines such neurons into layers. The input layer is an interface between the neural network and training dataset. Necessarily, it needs to have as many...