The currently implemented neural network ant has three user programmable parts:
Creation of the feature vector
Definition of output (actions)
Rewarding (Reinforcement learning)
Provide these parts as seperate functions for the user to write code for. Some additional settings could be needed (hidden layers, activation function etc.)
The currently implemented neural network ant has three user programmable parts:
Provide these parts as seperate functions for the user to write code for. Some additional settings could be needed (hidden layers, activation function etc.)