FEAT: Add tol option to best response related methods #74
Conversation
|
Thanks @QBatista
|
|
|
||
| # Returns | ||
| - `ne::Vector{NTuple{N,Int}}`: Vector of pure action Nash equilibria. | ||
| """ | ||
| function pure_nash(nfg::NormalFormGame; ntofind=Inf) | ||
| function pure_nash(nfg::NormalFormGame; ntofind=Inf, tol::Float64=1e-8) |
Codecov Report
@@ Coverage Diff @@
## master #74 +/- ##
==========================================
+ Coverage 94.28% 95.15% +0.87%
==========================================
Files 5 5
Lines 350 351 +1
==========================================
+ Hits 330 334 +4
+ Misses 20 17 -3
Continue to review full report at Codecov.
|
src/normal_form_game.jl
Outdated
| @@ -306,7 +306,7 @@ end | |||
| """ | |||
| best_response(player, opponents_actions, payoff_perturbation) | |||
|
|
|||
| Return the perturbed best response to `opponents_actions`. | |||
| Return the perturbed best responses to `opponents_actions`. | |||
src/normal_form_game.jl
Outdated
| throw(ArgumentError( | ||
| "tie_breaking must be one of 'smallest' or 'random'" | ||
| )) | ||
| end |
There was a problem hiding this comment.
I don't think we need these tie_breaking and tol for this method. payoff_perturbation is supposed to break ties (assuming that perturbations are drawn from some continuous distribution).
There was a problem hiding this comment.
I did this to match the Python version which has both. Do you want me to get rid of it?
src/normal_form_game.jl
Outdated
|
|
||
| # Returns | ||
|
|
||
| - `::Bool` | ||
| """ | ||
| is_nash(g::NormalFormGame{1}, action::Action) = | ||
| is_best_response(g.players[1], action, nothing) | ||
| is_nash(g::NormalFormGame{1}, action::Action, tol::Float64=1e-8) = |
src/normal_form_game.jl
Outdated
| is_nash(g::NormalFormGame{1}, action::Action, tol::Float64=1e-8) = | ||
| is_best_response(g.players[1], action, nothing, tol=tol) | ||
|
|
||
| is_nash(g::NormalFormGame{1}, action_profile::ActionProfile) = |
|
@oyamad This should be ready -- do you want me to squash into a few logical commits before reviewing it? |
|
Yes, please. |
Close #21 #75