/dnn_course

Solutions for Deep Neural Networks course at University of Warsaw.

Primary LanguageJupyter NotebookMIT LicenseMIT

dnn_course

Solutions for Deep Neural Networks course at University of Warsaw, 2021.

repo structure

Each jupyter notebook corresponds to different homework. Every homework received the maximum number of points.

tasks

  1. Verifying impact of data noise level with simple MLP architecture for MNIST dataset - code

  2. Object detection using anchor-based approach. Model's architecture is the variant of RetinaNET. Dataset consisted of images with MNIST digits (1-5), randomly flipped by 90 degrees - code

  3. LSTM Sentiment Tagger for imdb reviews dataset - code

  4. Implementation of DDPG for solving 3 different environments: Pendulum-v1, LunarLanderContinuous-v2, MountainCarContinuous-v0 - code