This repository contains the code for Categorical Naive Bayes Maximum Likelihood (MLE) and Maximum A-Posteriori (MAP) Estimators for the EMNIST Dataset.
For MAP, we assume a Dirichlet distribution for the class priors and a Beta distribution for the pixel priors.
We plot learning curves for MLE and various values of the MAP hyper-parameters and compare the results.
Full problemset for the CS-535 assignment available here.