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
Accept rational-precision mixed profiles for liap_solve #721
At the moment liap_solve expects the strategy profile passed to be of float (double) precision (MixedStrategyProfileDouble). It would be good to allow a rational-precision profile too, converting it to a double profile internally within the solver function call.
This is completely analogous to issue #458 which is for logit_estimate.
changed the title [-]Accept rational-precision mixed profiles for logit_estimate[/-][+]Accept rational-precision mixed profiles for `liap_solve`[/+]on Aug 6, 2026
@rahulsavani What would your view be on adding a .as_double() conversion to the mixed (strategy|behavior) profile instead? This would make it straightforward to convert, but require the user to convert.
Of course that is not mutually exclusive to accepting any profile as a convenience.
Within reason I support convenience, so tend to think we should do both, adding the conversion to the profile, and allowing any profile in the solvers as a convenience.
Activity
[-]Accept rational-precision mixed profiles for logit_estimate[/-][+]Accept rational-precision mixed profiles for `liap_solve`[/+]tturocy commentedon Aug 6, 2026
@rahulsavani What would your view be on adding a
.as_double()conversion to the mixed (strategy|behavior) profile instead? This would make it straightforward to convert, but require the user to convert.Of course that is not mutually exclusive to accepting any profile as a convenience.
rahulsavani commentedon Aug 23, 2026
Within reason I support convenience, so tend to think we should do both, adding the conversion to the profile, and allowing any profile in the solvers as a convenience.
Add `.as_float` to convert mixed profiles to floating-point equivalen…