Hi there 👋

Here is a list of ongoing and past projects I've been working on during my free time.

Ongoing projects

gym-routing [rl, graphs]: Solving the Traveling Salesman Prolem with reinforcement-learning using gym and stable-baselines3 repo
simple-implementations [llm, computer-vision]: Implementing ML and DL concepts (Transformer, LoRA, NeRF) from scratch repo

Previous Projects

nim-rl [rl]: Using numpy only, training an agent to master the Nim Game repo
Frequency-GANs [gan]: Generate market-specific financial time series using GANs acting in the frequency domain repo
Droughts Forecasting [lstm]: Forecast droughts in the US with a 6-week lead time using weather time series and static soil data repo
Image Completion [gan]: This model can effectively replace a nose or an eye on faces, and fill missing parts of portraits repo
Dynamic Pricing Optimization [ml, optimization]: Optimize Real Estate prices dynamically to meet financial constraints and deadlines (loan repayment dates) and maximize revenues repo

pierrecounathe's Projects

deep-rl-class

This repo contains the syllabus of the Hugging Face Deep Reinforcement Learning Course.

droughts-forecasting

Predicting droughts with a 6-weeks lead time with a lstm, using weather data, soil data.

financial-time-series-generation-frequency-gans

Generation of financial time series using GANs. From QuantGAN, TAGAN, and TTGAN, devised generators and discriminators acting in the frequency domain instead of the time domain. The time domain discriminator allows for better replication of autocorrelation patterns.

globally-and-locally-consistent-image-completion-pytorch-implementation

Implementation of the Globally and Locally Consistent Image Completion paper (http://dx.doi.org/10.1145/3072959.3073659) in Pytorch

gym-routing

Gym-Routing offers a suite of Gymnasium environments tailored for training Reinforcement Learning (RL) agents to tackle various routing problems.

gymnasium

An API standard for single-agent reinforcement learning environments, with popular reference environments and related utilities (formerly Gym)

nba-games-visualization-and-shooting-success-probability

This Sports Analytics project aims at creating a visualization tool for basketball games, incorporating the shooting success probability at any time.

pierrecounathe.github.io

This repo contains my personal website

real-estate-pricing-optimization-under-financial-constraints

This repository contains a high-level description of a Pricing Optimization project in Real Estate (RE). The scope of this project ranges from data scraping, to single-point market value prediction for RE assets, to time-on-market or time-to-sell prediction given a price, to price optimization.

reinforcement-learning-for-nim-game

The Nim game is a mathematical strategy game in which two players take turns removing sticks from distinct heaps. The player who removes the last stick loses. As a game in which states are fully observable and with much reduced state and action spaces compared to chess, it is the perfect game to code a Q-Learning RL agent from scratch.