dLife’s NN / GA / RL Packages
Neural Networks
The dLife neural network package provides support for multilayer feedforward networks, CMAC networks and Elman networks. Backpropagation training is supported for multilayer feedforward networks and for Elman networks. Creating and training networks requires only a few lines of code. Trained networks are easily saved to and loaded from files.
Genetic Algorithms
The dLife genetic algorithms package provides support for generational genetic algorithms. A variety of gene types, reproductive operators and selection mechanisms are included. For more advanced applications the package design facilitates the addition of new types of genes, reproductive operators and selection mechanisms.
Reinforcement Learning
The dLife reinforcement learning package provides support for a variety of TD based RL policies including Q-Learning, R-Learning and SARSA. The user configures and assembles objects representing possible actions, a Q-value update policy, an action selection policy, and the learning rate. The provided algorithms do the rest. Additional polices for action selection, Q-value updates and learning rates are easily added.
