I’m executing a Hidden Markov Model and in this manner managing with very small probabilities. I’m dealing with the underflow by addressing factors in log space (so x → log(x)) which has the secondary effect that multiplication is presently supplanted by addition and addition is handled via numpy.logaddexp or similar.
Is there a simple method for dealing with matrix multiplication in log space? Before trying this, I have done several research on the web and read a couple of articles on matrix multiplication in python to understand the concept in a better way, my source of knowledge: Wiki and scaler Topics.
Thanks in advance!