Local online learning in recurrent networks with random feedback
Abstract
Recurrent neural networks (RNNs) enable the production and processing of time-dependent signals such as those involved in movement and working memory. Classic gradient-based algorithms for training RNNs have been available for decades, but are inconsistent with biological features of the brain, such as causality and locality. We derive an approximation to gradient-based learning that comports with these constraints by requiring synaptic weight updates to depend only on local information about pre- and postsynaptic activities, in addition to a random feedback projection of the RNN output error. In addition to providing mathematical arguments for the effectiveness of the new learning rule, we show through simulations that it can be used to train an RNN to perform a variety of tasks. Finally, to overcome the difficulty of training over very large numbers of timesteps, we propose an augmented circuit architecture that allows the RNN to concatenate short-duration patterns into longer sequences.
Data availability
Code implementing the RFLO learning algorithm for the example shown in Figure 2 has been included as a source code file accompanying this manuscript.
Article and author information
Author details
Funding
National Institutes of Health (DP5 OD019897)
- James M Murray
National Science Foundation (DBI-1707398)
- James M Murray
Gatsby Charitable Foundation
- James M Murray
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Copyright
© 2019, Murray
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 6,054
- views
-
- 951
- downloads
-
- 53
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Neuroscience
Intrinsic sensory neurons are an essential part of the enteric nervous system (ENS) and play a crucial role in gastrointestinal tract motility and digestion. Neuronal subtypes in the ENS have been distinguished by their electrophysiological properties, morphology, and expression of characteristic markers, notably neurotransmitters and neuropeptides. Here, we investigated synaptic cell adhesion molecules as novel cell-type markers in the ENS. Our work identifies two type II classic cadherins, Cdh6 and Cdh8, specific to sensory neurons in the mouse colon. We show that Cdh6+ neurons demonstrate all other distinguishing classifications of enteric sensory neurons including marker expression of Calcb and Nmu, Dogiel type II morphology and AH-type electrophysiology and IH current. Optogenetic activation of Cdh6+ sensory neurons in distal colon evokes retrograde colonic motor complexes (CMCs), while pharmacologic blockade of rhythmicity-associated current IH disrupts the spontaneous generation of CMCs. These findings provide the first demonstration of selective activation of a single neurochemical and functional class of enteric neurons and demonstrate a functional and critical role for sensory neurons in the generation of CMCs.
-
- Evolutionary Biology
- Neuroscience
Genetic variation is known to contribute to the variation of animal social behavior, but the molecular mechanisms that lead to behavioral differences are still not fully understood. Here, we investigate the cellular evolution of the hypothalamic preoptic area (POA), a brain region that plays a critical role in social behavior, across two sister species of deer mice (Peromyscus maniculatus and P. polionotus) with divergent social systems. These two species exhibit large differences in mating and parental care behavior across species and sex. Using single-nucleus RNA-sequencing, we build a cellular atlas of the POA for males and females of both Peromyscus species. We identify four cell types that are differentially abundant across species, two of which may account for species differences in parental care behavior based on known functions of these cell types. Our data further implicate two sex-biased cell types to be important for the evolution of sex-specific behavior. Finally, we show a remarkable reduction of sex-biased gene expression in P. polionotus, a monogamous species that also exhibits reduced sexual dimorphism in parental care behavior. Our POA atlas is a powerful resource to investigate how molecular neuronal traits may be evolving to give rise to innate differences in social behavior across animal species.