A solution to the learning dilemma for recurrent networks of spiking neurons

Recurrently connected networks of spiking neurons underlie the astounding information processing capabilities of the brain. Yet in spite of extensive research, how they can learn through synaptic plasticity to carry out complex network computations remains unclear. We argue that two pieces of this p...

Full description

Saved in:
Bibliographic Details
Published in:Nature communications 2020-07, Vol.11 (1), p.3625-3625, Article 3625
Main Authors: Bellec, Guillaume, Scherr, Franz, Subramoney, Anand, Hajek, Elias, Salaj, Darjan, Legenstein, Robert, Maass, Wolfgang
Format: Article
Language:English
Subjects:
631/114/116/1925
631/378
631/378/116/2396
631/378/2591
639/166/987
Artificial intelligence
Back propagation
Back propagation networks
Data processing
Energy efficiency
Firing pattern
Hardware
Humanities and Social Sciences
Information processing
Learning algorithms
Machine learning
Mathematical models
multidisciplinary
Nervous system
Neural networks
Neurons
Recurrent neural networks
Science
Science (multidisciplinary)
Short term memory
Spikes
Spiking
Synaptic plasticity
Training
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Recurrently connected networks of spiking neurons underlie the astounding information processing capabilities of the brain. Yet in spite of extensive research, how they can learn through synaptic plasticity to carry out complex network computations remains unclear. We argue that two pieces of this puzzle were provided by experimental data from neuroscience. A mathematical result tells us how these pieces need to be combined to enable biologically plausible online network learning through gradient descent, in particular deep reinforcement learning. This learning method–called e-prop–approaches the performance of backpropagation through time (BPTT), the best-known method for training recurrent neural networks in machine learning. In addition, it suggests a method for powerful on-chip learning in energy-efficient spike-based hardware for artificial intelligence. Bellec et al. present a mathematically founded approximation for gradient descent training of recurrent neural networks without backwards propagation in time. This enables biologically plausible training of spike-based neural network models with working memory and supports on-chip training of neuromorphic hardware.
ISSN:2041-1723
2041-1723
DOI:10.1038/s41467-020-17236-y