In this repo you'll find all four home assignments I completed, as part of the Deep Learning course at the Ben-Gurion University of the Negev. Each task covers one of the following fields: ANN, CNN, RNN, and GAN.
The course begins with an overview of the underlying theories in the field of DL, then reviewing the many current network types and their applicability to diverse applications. The course's introduction covered the whole syllabus of Andrew Ng's "DeepLearning Specialization" course on Coursera (link).
Assignment 1 (ANN) - In this task, we implemented a straightforward neural network from scratch (without utilizing TensorFlow or Keras), to understand the processes of forward/backward propagation.
Assignment 2 (CNN) - Facial Recognition using One-shot Learning. In this assignment, we performed a one-shot learning task for previously unseen objects. We used the architecture described in the paper Siamese Neural Networks for One-shot Image Recognition. The goal: Given two facial images of previously unseen persons, determine whether they are the same person.
Assignment 3 (RNN) - lyrics generation- In this task, we trained a RNN to generate lyrics based on a provided melody. We were given .midi files that contain various types of information such as notes, or the instruments used. We were challenged to incorporate this information with the lyrics and automatically generate lyrics for a provided melody.
Assignment 4 (GAN) - Tabular samples generation - In this task, we implemented and explored two GAN architectures for generating tabular data.
