This repository is intended for the final project of the course '42186 - Model-based Machine Learning' from the Technical University of Denmark (DTU). It contains all the data, scripts, literature, and final report.
All assignments are coded in Python using Pyro for the probabilistic programming and JAX for the speed-ups advantages.
The task is based on the Kaggle competition "Great Energy Predictor III", which contains a dataset with hourly measurements of energy consumption on different types of buildings in different sites (or regions). The imlplemented algorithms were: vanilla linear regression, bayesian linear regression, hierarchical regression (applied to a bunch of households), and several Linear Dynamical Systems with lags 1, 2, 24, and 168.