A Generalized Framework for Self-Play Training
Win or Learn Fast Proximal Policy Optimisation
TrueMatch Matchmaking System
The TrueMatch matchmaking system decides which people should play together in an online multiplayer game. TrueMatch was developed by Microsoft Research and launched in Gears 5. It was subsequently incorporated into Gears of War 4…
Time discretization invariance in Machine Learning, applications to reinforcement learning and recurrent neural networks
While computers are well equipped to deal with discrete flows of data, the real world often provides intrinsically continuous time data sequences, e.g. visual, sensory streams, time series, or state variables in continuous control environments.…
Sequential Estimation of Quantiles with Applications to A/B-testing and Best-arm Identification
Consider the problem of sequentially estimating quantiles of any distribution over a complete, fully-ordered set, based on a stream of i.i.d. observations. We propose new, theoretically sound and practically tight confidence sequences for quantiles, that…
Towards Secure and Interpretable AI: Scalable Methods, Interactive Visualizations, and Practical Tools
The explosion of available idea repositories — scientific papers, patents, product descriptions — represents an unprecedented opportunity to accelerate innovation and lead to a wealth of discoveries. Given the scale of the problem and its…
Interpretable Outcome Prediction with Sparse Bayesian Neural Networking
A type of Bayesian Neural Network which has a sparsity-inducing prior distribution in order to help interpret the learned weights. Particularly useful in the domain of healthcare but scales to many other domains, as it…