You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug]: enumpoly hanging on 6x6 bimatrix game #1055
The bug is shown in the following example code where we load a 6x6 bimatrix game with 75 equilibria from the catalog.
Using enummixed_solve we find all 75 equilibria, but in the below code with enumpoly_solve, after creating an EFG version of the bimatrix game we then struggle to find the:
54 equilibria with use_strategic=True (but finding 53 is fine);
1 equilibrium with with use_strategic=False.
We find the same with the "small payoff" version of this game (slug: "6x6_game_with_75_eq_small_payoffs").
import pygambit as gbt
from tests.games import create_efg_corresponding_to_bimatrix_game
slug = "journals/dcg/vonstengel1999/6x6_game_with_75_eq"
A, B = gbt.catalog.load(slug).to_arrays()
g = create_efg_corresponding_to_bimatrix_game(A, B, "")
#######################
# use_strategic is True
#######################
# Runs in ~ 10 seconds:
print(gbt.nash.enumpoly_solve(g, stop_after=53, use_strategic=True))
# Hangs:
# print(gbt.nash.enumpoly_solve(g, stop_after=54, use_strategic=True))
########################
# use_strategic is False
########################
# Hangs:
# print(gbt.nash.enumpoly_solve(g, stop_after=1, use_strategic=False))
Activity
Improve handling of less-well-conditioned systems in `enumpoly_solve`.
Improve handling of less-well-conditioned systems in `enumpoly_solve`.
Improve handling of less-well-conditioned systems in `enumpoly_solve`.
Improve numerical stability and robustness of `enumpoly` (gambitproje…