gregora / Cpp-neural-network

A neural network class, written in C++ (LEGACY)

C++ neural network

A neural network class, written in C++

How to use this network

Create your own programm and include "network.h" header file with "network.cpp" file. To create an instance of class run Network network(x, y, z, n);. X here represents the number of input nodes, y number of hidden layers, z number of nodes in each hidden layer and n number of output nodes.

Functions

• network.random() returns a double (a random number between 0 and 1)
• network.sigmoid(double in) is an approximation to sigmoid function. Accepts double argument and squishes it to a number between -1 and 1
• network.set_input_nodes(double nodes[]) accepts an array of values, that input nodes should be set to. This function sets all the other nodes to 0
• network.reset_nodes() sets all the nodes to 0
• network.reset_edges() sets all the edges to 0
• network.mutate(int amount) changes a random edge to a random number from 0 to [chosen] amount
• network.loss(double expected_output[]) accepts and array of what output nodes should be like and compares it to the actual values and returns loss
• network.forward_propagate() propagates forward and changes all the values of nodes
• network.output_all_nodes() outputs all node values (mainly used for debugging)
• network.save(std::string file_path) saves the network to a file
• network.load(std::string file_path) loads a network from a file (note: this only changes edges and not nodes)

visualize.py

This script was created, to more easily visualize a saved network. It is written in Python 2, and all it does, is read from a saved network, and renders it in a window.