The final project of the course AMNLP (Advanced methods in NLP). Shimon Malnick and I tried to solve the CR resolution task on the OntoNotes benchmark by fine-tuning BART and T5.

Pre-trained language models (LMs) have recently demonstrated outstanding results across a variety of tasks. However, it remains unclear precisely what knowledge the LM manages to capture during pre-training and how word frequency in the training corpus affects the acquisition of knowledge about these words.

As part of of my passion for CV algorithms, I challenged myself to implement a real-time algorithm for face detection from scratch. The Viola Jones is a very light-weight FD algorithm that combines a set of weak (and fast) classifiers into a single strong classifier using boosting.

A simple reinforcement learning (snake) game that I wrote for fun while exploring RL during my Masters. The learning algorithm of the snake is based on Monte-Carlo using every visit update.