Open jbendelac opened 5 years ago
Hi. Good questions. Could you elaborate a little more? If you're asking about "discrete optimization", we didn't cover any of this in class. If you're wondering about using MLE to estimate the parameters of discrete random variables, such estimation falls within the domain of "continuous optimization." Our usual estimation procedures should work for this. Though, there are a lot of times when an optimizer will fail for one reason or another. If you're trying to optimize a function of one variable, sometimes using scipy.optimize.minimize_scalar
works better.
I'm wondering about using MLE to estimate parameters of discrete random variables. My probability mass function and neg log lik functions work correctly, but my opt.minimize is returning the initial guess, no matter what the initial guess is. How would I go about fixing this problem?
Ok. Have you tried scipy.optimize.minimize_scalar
? Also, have you tried other optimization methods within minimize
? Take a look at the method
keyword argument: https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html
But try minimize_scalar
first.
Hello, I'm practicing running mle on discrete functions, and have been having trouble using the opt.minimize function with discrete datasets in the examples I am working with. Can opt.minimize work with discrete data, and if not is there another function I can use instead of opt.minimize?