I've added a skip_check argument (set to False by default) in MDP and all subclasses. I think it might be useful to skip checking the transition and reward matrices on the creation of every MDP instance if one knows that they are correct beforehand (which can be verified using mdptoolbox.util.check). For example, this can save some time when the same MDP is solved using different algorithms.
Coverage increased (+0.04%) to 87.12% when pulling ffaef478fd3ad63b355b0d6a17ee0c7541a650b2 on yasserglez:skip_check into 8fac9a11e0792606eef15b1df450307e7f9d6a8c on sawcordwell:master.
Sounds good. The Travis build error is just that pip couldn't download cvxopt on one of the tests, but the others all pass so I'm confident that it is working as expected.
I've added a
skip_check
argument (set toFalse
by default) in MDP and all subclasses. I think it might be useful to skip checking the transition and reward matrices on the creation of every MDP instance if one knows that they are correct beforehand (which can be verified usingmdptoolbox.util.check
). For example, this can save some time when the same MDP is solved using different algorithms.