Training artificial neural networks to learn a nondeterministic game
This addresses the challenge of enabling ANNs to handle unpredictable real-world phenomena, but appears incremental as it applies existing models to a new nondeterministic scenario.
The paper tackled the problem of training artificial neural networks to learn nondeterministic automata, using the game of Pong as a test environment with incomplete state information, and tested three models (Mona, Elman, NuPIC) to assess their performance.
It is well known that artificial neural networks (ANNs) can learn deterministic automata. Learning nondeterministic automata is another matter. This is important because much of the world is nondeterministic, taking the form of unpredictable or probabilistic events that must be acted upon. If ANNs are to engage such phenomena, then they must be able to learn how to deal with nondeterminism. In this project the game of Pong poses a nondeterministic environment. The learner is given an incomplete view of the game state and underlying deterministic physics, resulting in a nondeterministic game. Three models were trained and tested on the game: Mona, Elman, and Numenta's NuPIC.