All about modular optimisation of robots, teach-in, independent test set, statistics, pattern recognition, self learning, machine learning, ID3, neural networks, genetic programing, koza, computer vision - and how to avoid overlearning.

Suitable applications for self learning algorithms include
  • humanoid robots
  • chess game
  • forex (algorithmic trading robots)
  • What makes these applications suitable for neural networks and other self learning algorithms ? How to find suitable learning samples and a truly independent test set ? What is the very best goal function ?