Open source
Libraries and tools I built.
High-throughput asynchronous reinforcement learning framework. At the time of release, the fastest open-source PPO implementation: ~10x faster than traditional synchronous RL implementations, with SOTA results in challenging VizDoom and DMLab environments. Agents trained with Sample Factory:
The fastest (at the time of release) embodied simulator for AI research: 1,000,000+ FPS of immersive, physics-based multi-agent experience on a single machine.
A faster alternative to Python's built-in multiprocessing.Queue.
Python implementation of the Qt-like signal & slot asynchronous programming paradigm, with focus on multiprocessing applications.
"4D video" grabber and player for Intel RealSense and Google Tango, with a fast real-time Delaunay triangulation (modified Guibas-Stolfi): 300fps on PC, 100fps on Android. More in other projects.
Smaller projects
- tf-reinforce: Tensorflow implementation of classic policy gradient algorithm for continuous control tasks.
- snake-rl: RL algorithms for classic Snake game (youtube).
- hyperopt.py: evolutionary algorithm for hyperparameter optimization in deep learning (it works!).
- udacity-linear-algebra-cpp: small linear algebra library in C++11/14 with templates, SFINAE, etc.