Timezone: »

Memory-Based Meta-Learning on Non-Stationary Distributions
Tim Genewein · Gregoire Deletang · Anian Ruoss · Li Kevin Wenliang · Elliot Catt · Vincent Dutordoir · Jordi Grau-Moya · Laurent Orseau · Marcus Hutter · Joel Veness

Wed Jul 26 02:00 PM -- 03:30 PM (PDT) @ Exhibit Hall 1 #511
Event URL: https://github.com/deepmind/nonstationary_mbml »

Memory-based meta-learning is a technique for approximating Bayes-optimal predictors. Under fairly general conditions, minimizing sequential prediction error, measured by the log loss, leads to implicit meta-learning. The goal of this work is to investigate how far this interpretation can be realized by current sequence prediction models and training regimes. The focus is on piecewise stationary sources with unobserved switching-points, which arguably capture an important characteristic of natural language and action-observation sequences in partially observable environments. We show that various types of memory-based neural models, including Transformers, LSTMs, and RNNs can learn to accurately approximate known Bayes-optimal algorithms and behave as if performing Bayesian inference over the latent switching-points and the latent parameters governing the data distribution within each segment.

Author Information

Tim Genewein (DeepMind)
Gregoire Deletang (DeepMind)
Anian Ruoss (Google DeepMind)
Li Kevin Wenliang (Google DeepMind)
Elliot Catt (DeepMind)
Vincent Dutordoir (University of Cambridge)
Jordi Grau-Moya (DeepMind)
Laurent Orseau (DeepMind)
Marcus Hutter (DeepMind)
Joel Veness (Google)

More from the Same Authors