Neural Networks
Last updated
Last updated
One of the most successful methods of ML.
Building blocks: Perceptron and Synapse.
Typically a function that maps the entire natural range into a bounded interval ([0,1] or [-1,1]).
Logistic sigmoid function ReLU function, tanh, softmax, etc.
Connections from perceptrons of layer (n-1)th to perceptrons of layer nth, each applying a weight to the transmitted value.
Training Neural Networks is mostly about finding the weights of those synapses.