A Neurobiologically-inspired Deep Learning Framework for Autonomous Context Learning

Abstract

Neurobiologically-inspired working memory models have managed to accurately demonstrate and explain our ability to rapidly adapt and alter our responses to the environment. However, the applications of these working memory models have been limited to reinforcement learning problems. Furthermore, the incorporation of contextual/switching mechanisms outside of the realm of working memory modeling for general-use cases has also been relatively unexplored. We present a new framework compatible with Tensorflow Keras enabling the straightforward integration of working memory-inspired mechanisms into typical neural network architectures. These mechanisms allow models to autonomously learn multiple tasks, statically or dynamically allocated. We also examine the generalization of the framework across a variety of multi-context supervised learning and reinforcement learning tasks. The resulting experiments successfully integrate these mechanisms with multilayer and convolutional neural network architectures. The diversity of problems solved demonstrates the framework’s generalizability across a variety of architectures and tasks.