| Paper: | PS-2A.5 | ||
| Session: | Poster Session 2A | ||
| Location: | H Lichthof | ||
| Session Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation: | Poster | ||
| Publication: | 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany | ||
| Paper Title: | Q-AGREL: Biologically Plausible Attention Gated Deep Reinforcement Learning | ||
| Manuscript: | Click here to view manuscript | ||
| License: |  This work is licensed under a Creative Commons Attribution 3.0 Unported License. |
||
| DOI: | https://doi.org/10.32470/CCN.2019.1243-0 | ||
| Authors: | Isabella Pozzi, Sander M. Bohté, Centrum Wiskunde & Informatica, Netherlands; Pieter R. Roelfsema, Netherlands Institute for Neuroscience, Netherlands | ||
| Abstract: | The success of deep learning in end-to-end learning on a wide range of complex tasks is now fuelling the search for similar deep learning principles in the brain. While most work has focused on biologically plausible variants of error-backpropagation, learning in the brain seems to mostly adhere to a reinforcement learning paradigm, and while biologically plausible neural reinforcement learning has been proposed, these studies focused on shallow networks learning from compact and abstract sensory representations. Here, we demonstrate how these learning schemes generalize to deep networks with an arbitrary number of layers. The resulting reinforcement learning rule is equivalent to a particular form of error-backpropagation that trains one output unit at anytime. We demonstrate the learning scheme on classical and hard image-classification benchmarks, namely MNIST, CIFAR10 and CIFAR100, cast as direct reward tasks, both for fully connected, convolutional and locally connected architectures. We show that our learning rule - Q-AGREL - performs comparably to supervised learning via error-backpropagation, requiring only 1.5-2.5 times more epochs, even when classifying 100 different classes as in CIFAR100. Our results provide new insights into how deep learning may be implemented in the brain. | ||