search

TheRealJP / Markov-Decision-Proces

0 stars 0 forks source link

Overflow error in Value Iteration #1

Open Nick-VO opened 5 years ago

Nick-VO commented 5 years ago

edb5e771cf778ee481c1f3f391f2bf64c3d82090

Utility values increase exponentially when a reward-giving state is found. The program looks like it's stuck for a few seconds then gives an overflow exception. The exception happens because the utility value surpass the float64 limit.

Error happens on line 21 in https://github.com/TheRealJonathanPeers/Skynet_alpha/blob/feature_ai/ai/evaluations/value_iteration.py

Nick-VO commented 5 years ago

found out error can be resolved by lowering the discount factor