Connor Watts's Projects
Trading bot with support for realtime trading, backtesting, custom strategies and much more.
Code for "Infinitely Deep Bayesian Neural Networks with Stochastic Differential Equations"
Personal research website
Code for the online course "Deployment of Machine Learning Models"
Fully differentiable RL environments, written in Ivy.
Reinforcement Learning (RL) environments for a range of graph problems. Written in Jraph.
Liquid Structural State-Space Models
MLOps examples
Multi-Task Learning framework for Image Generative Adversarial Network
Multi-task learning models built for song genre classification made as a group project for the Natural Language Processing course at UCL. Here we consider both BERT and CNN based architectures.
A multi-task learning model built for image segmentation made as a group project for the Deep Learning course at UCL
Partially Stochastic but Infinitely Deep Neural Networks
Reinforcement Learning based Portfolio Manager. Here the agent allocate portions of a fund to different financial products with the goal of maximising returns. Currently supports DDPG and PPO.
Deep Reinforcement Learning based Rubikโs Cube solver written in JAX, Haiku and RLax. Developed a custom environment supporting cubes of variable size. Currently developing support for DQN, QR-DQN and PPO.
Two Generalized Energy Based Models for time-series generation. Thesis for MSc Computational Statistics and Machine Learning program at UCL.
Differentiable SDE solvers with GPU support and efficient sensitivity analysis.