The parallels between statistical mechanics and machine learning are eerily profound. Optimization theory describes systems and their evolution toward lowest-en

The parallels between statistical mechanics and machine learning are eerily profound. Optimization theory describes systems and their evolution toward lowest-energy states, the landscape that optimization traverses and the dynamics by which it does so are described by statistical mechanical principles. I had an experience in an interview for a Sr MLE role describing some previous work. Was unable to get through to the interviewer that the answer to the question of "which loss function did you use" can't always be found in a Medium listicle. Hint: loss functions are negative log-likelihoods.