Distributions, expectation, Bayes and maximum likelihood — the vocabulary every loss function is written in.
10h · 7 lessons · 1 challenge
Cross-entropy is a likelihood. Regularization is a prior. Dropout is a random variable. This module gives you the probability to see loss functions as the statistical statements they are, instead of formulas to memorize.
It ends with maximum likelihood estimation, which is where the maths stops being background and starts being the derivation of every objective you will use for the rest of the course.
Rapid-fire posterior estimates against a clock and against other learners. Points for calibration, not just for being right.
Worth 350 XP.
