search

shanive / bgu-ailab-bandits

bandits expiriment
6 stars 0 forks source link

implement tournament for several agents #25

Closed dtolpin closed 13 years ago

dtolpin commented 13 years ago

implement tournament for several agents, all-to-all n(n-1)*number-of-repetitions for now. allow to specify multiple agents on the command-line: for example agents are positional arguments, other args are key arguments.

For example,

tournament --size 20 --repeat 1000 --samples 100 uct uniform random

means 20 switches, 100 samples per node in Monte Carlo sampling algorithms, repeat each experiment 1000 times

shanive commented 13 years ago

Isn't is more suitable for play.py program? Because of the single size instead of min-size and max-size in tournament(sosTest.py).