seqlearn icon indicating copy to clipboard operation
seqlearn copied to clipboard

Constraints on Transition Probabilities

Open closedLoop opened this issue 9 years ago • 0 comments

I needed the ability to put constraints on some transition probabilities. Namely I had to set a few transition probabilities close to 0. Here is an initial attempt which uses numpy.ma.masked_array to fix some of the values in b_trans.

I utilize a hyper-parameter named trans_constraints to provide a list of transitions that should have very low probabilities.

I have provided a simple test to show that the constraints work.

Any feedback would be appreciated.

closedLoop avatar Feb 20 '16 03:02 closedLoop