With a large model and multiple transitions, the probability can “underflow,” becoming indistinguishable from zero as represented with the floating point numbers used by MATLAB. One way to circumvent this is to convert take the logarithm of the probability and work in a logarithmic scale, converting the value back before returning. Update the above function to calculate the probabilities as logarithms. You will need to address zero-valued probabilities. This may be as simple as using negative infinity (in MATLAB, log(0) = -Inf).
Already registered? Login
Not Account? Sign up
Enter your email address to reset your password
Back to Login? Click here