Tag: maximum likelihood

• Logistic Regression: Deriving the Sigmoid and the Cross Entropy from Maximum LikelihoodThe sigmoid and the logit turn a linear score into a probability, and Bernoulli maximum likelihood turns into the cross entropy error; the gradient and Hessian show the loss is convex but has no closed form.Computer ScienceMathematics for Machine LearningUndergraduatelogistic regressionsigmoid functioncross entropymaximum likelihoodconvex optimisation~28 min
• Why Machine Learning Needs Mathematics: Rewriting Learning as Loss MinimisationRegression and classification both reduce to a single optimisation problem, empirical risk minimisation. We show where linear algebra, calculus and probability enter, with complete proofs for least squares and gradient descent.Computer ScienceMathematics for Machine LearningUndergraduatemachine learningempirical risk minimisationleast squaresgradient descentmaximum likelihood~23 min