This is the implementation of an assignment from the Master of Science's course "Artificial Neural Networks and Deep Learning" of Politecnico di Milano. The project was done in a group of 3, in collaboration with my colleagues Davide Mantegazza and Gabriele Bozzetto.
In this homework, we were required to predict future samples of a multivariate time series like the one shown in the image below. The goal is to design and implement forecasting models to learn how to exploit past observations in the input sequence to correctly predict the future.
We implemented various models and explored aspects such as the LSTM architecture, hyper-parameters tuning, the attention mechanism, and the transformer-based models.
Each notebook contains a particular model, with images of the architecture and its RTSE on the validation set. Further details are found in the "AN2DL HW02 report.pdf" file.
