/Deep-Learning-MVA

Deep Learning MVA

Primary LanguageJupyter Notebook

This repository contains my homeworks for the Deep learning course of the MVA master.

Class website: https://www.labri.fr/perso/vlepetit/deep_learning_mva.php

We had to do three homeworks:

HW1: simple regression and classification task (classify pictures (between the classes circle, triangle and rectangle), given a picture of a triangle given the vertices coordinates, denoise some simple pictures using hourglass networks).

HW2: NLP based project (multilingual word embeddings using an alignment matrix, sentence classification with BoV, sentence classification using recurrent neural network architectures).

HW3: Deep reinforcement learning (build an agent that will solve the following problem: A rat runs on an island and tries to eat as much as possible. The island is subdivided into $N\times N$ cells, in which there are cheese (+0.5) and poisonous cells (-1). The rat has a visibility of 2 cells (thus it can see $5^2$ cells). The rat is given a time $T$ to accumulate as much food as possible. It can perform 4 actions: going up, down, left, right).