For better understand of neural networks I started implementation of Multi Layer Perceptron. For now I'm implemented single Perceptron that resolve XOR problem. From this point I want start build MLP but I'm not sure if I correctly understand MLP structure. Assume I have data instance with 3 attributes and 2 classes and 5 input perceptrons. Should I give every perceptron same input and every output of perceptron should go to every hidden layer perceptron's input? I attach neural net pic.
With MLPs you can follow a set of simple rules:
- The number of neurons in your input layer equals the number of features / data instances you have.
- A single neuron takes only one input (one feature)
- The output from one neuron can go to multiple neurons in the next layer which may or may not be fully connected.
May the force be with you ;)