Added support for the Adadelta training algorithm, a neural network pruning algorithm (Karnin 1990), and a robotic simulator (though currently undocumented). More tests were written for functionality across the library. We made class structure changes to make it easier to add new functionality to the Fido library.

v0.0.1 - First release

• Neural networks
  • Multilayer feed-forward neural networks
  • Sigmoid, linear, and hyperbolic tangent activation functions
• Trainers
  • SGD backpropagation
  • Adadelta
• Reinforcement Learning
  • Q-learning
  • Wire-fitted q-learning
  • A universal robot control system
• Genetic algorithms