
AI Blueprints
By :

Neural networks, also known as artificial neural networks, are an ML paradigm inspired by animal neurons. A neural network consists of many nodes, playing the role of neurons, connected via edges, playing the role of synaptic connections. Typically, the neurons are arranged in layers, with each layer fully connected to the next. The first and last layers are input and output layers, respectively. Inputs may be continuous (but often normalized to [-1, 1]) or binary, while outputs are typically binary or probabilities. The network is trained by repeatedly examining the training set. Each repetition on the full training set is called an "epoch." During each epoch, the weights on each edge are slightly adjusted in order to reduce the prediction error for the next epoch. We must decide when to stop training, that is, how many epochs to execute. The resulting learned "model" consists of the network topology as well as the various weights.
Each neuron has a set...