Machine Learning Developer in London, United Kingdom

QMIX is a state-of-the-art value-based algorithm for collaborative deep multi-agent reinforcement learning for the settings of centralized training with decentralized execution.
QMIX employs a neural network that estimates joint action-values as a complex nonlinear combination of per-agent values that condition only on local observations. We structurally enforce that the joint-action value is monotonic in the per-agent values, which allows tractable maximization of the joint action-value in off-policy learning, and guarantees consistency between the centralized and decentralized policies.

QMIX was presented at ICML 2018, one of the leading conferences for machine learning research.

PyMARL is a research platform for deep multi-agent reinforcement learning which allows out-of-the-box experimentation and development.

Written in PyTorch, PyMARL features implementations of several state-of-the-art methods such as QMIX, COMA, and independent Q-Learning.

In collaboration with the Berkeley AI Research lab, some of the above algorithms have also been successfully ported to the scalable RLlib framework.

The StarCraft Multi-agent Challenge (SMAC) is a benchmark for cooperative multi-agent reinforcement learning (MARL) that provides elements of partial observability, challenging dynamics, and high-dimensional observation spaces. SMAC is built using the StarCraft II game engine, creating a testbed for research in cooperative MARL where each game unit is an independent RL agent.

Paper: https://arxiv.org/abs/1902.04043
Blogpost: http://whirl.cs.ox.ac.uk/blog/smac/
Blogpost by NVIDIA: https://nvda.ws/2I88F1I

SMAC was presented at AAMAS 2019, the largest and most influential conference in the area of agents and multiagent systems.

Our learning process depends on having a fairly rich (though developmentally plausible) input representation. This repository explores how two aspects of visual perception which are vital for early language learning can be captured by algorithms, namely motion cause detection and gaze object detection.