date | title | id | tags | ||
---|---|---|---|---|---|
2021-04-26 11:14:48 -0700 |
Artificial Neural Networks and Deep Learning |
2021-04-26t18-14-48z |
|
An Artificial Neural Network (ANN) is a kind of computing system (computes outputs based on inputs) inspired by the brain. In particular, an ANN is an arrangement of artificial neurons (units), with each individual neuron potentially being connected to more than one other neuron.
Typically, the units of an ANN are arranged in layers (not always, see GNNs). The input vector will form the first layer, and the output vector will form the final layer. In between we have what are referred to as "hidden layers".
Very simply put, the activations in one layer affect the activations in the next layer, until we finally obtain the output vector.
As such, a Neural Network can be thought of as a function, taking some inputs, and computing some outputs based on them.
Neural Networks are typically "deep", i.e. consisting in more than one hidden layer. For this reason, they are often directly referred to as "deep neural networks" and form the Machine Learning subfield that is referred to as Deep Learning.