Gated recurrence enables simple and accurate sequence prediction in stochastic, changing, and structured environments
Abstract
From decision making to perception to language, predicting what is coming next is crucial. It is also challenging in stochastic, changing, and structured environments; yet the brain makes accurate predictions in many situations. What computational architecture could enable this feat? Bayesian inference makes optimal predictions but is prohibitively difficult to compute. Here, we show that a specific recurrent neural network architecture enables simple and accurate solutions in several environments. This architecture relies on three mechanisms: gating, lateral connections, and recurrent weight training. Like the optimal solution and the human brain, such networks develop internal representations of their changing environment (including estimates of the environment's latent variables and the precision of these estimates), leverage multiple levels of latent structure, and adapt their effective learning rate to changes without changing their connection weights. Being ubiquitous in the brain, gated recurrence could therefore serve as a generic building block to predict in real-life environments.
Data availability
This paper presents no experimental data. All synthetic data are available in the code repository at https://github.com/cedricfoucault/networks_for_sequence_prediction and archived on Zenodo with DOI: 10.5281/zenodo.5707498.
-
Networks for sequence predictionPublicly available at Github (https://github.com).
-
Networks for sequence predictionPublicly available at Zenodo (https://zenodo.org/).
Article and author information
Author details
Funding
Ecole normale superieure Paris-Saclay (PhD fellowship Contrat doctoral spécifique normalien"")
- Cédric Foucault
Agence Nationale de la Recherche (18-CE37-0010-01 CONFI LEARN"")
- Florent Meyniel
H2020 European Research Council (ERC StG 947105 NEURAL PROB"")
- Florent Meyniel
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Copyright
© 2021, Foucault & Meyniel
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
-
- 1,246
- views
-
- 188
- downloads
-
- 6
- 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
-
- Cell Biology
- Neuroscience
Overactivity of the sympathetic nervous system is a hallmark of aging. The cellular mechanisms behind this overactivity remain poorly understood, with most attention paid to likely central nervous system components. In this work, we hypothesized that aging also affects the function of motor neurons in the peripheral sympathetic ganglia. To test this hypothesis, we compared the electrophysiological responses and ion-channel activity of neurons isolated from the superior cervical ganglia of young (12 weeks), middle-aged (64 weeks), and old (115 weeks) mice. These approaches showed that aging does impact the intrinsic properties of sympathetic motor neurons, increasing spontaneous and evoked firing responses. A reduction of M current emerged as a major contributor to age-related hyperexcitability. Thus, it is essential to consider the effect of aging on motor components of the sympathetic reflex as a crucial part of the mechanism involved in sympathetic overactivity.
-
- Cell Biology
- Neuroscience
Chronic high-fat feeding triggers metabolic dysfunction including obesity, insulin resistance, and diabetes. How high-fat intake first triggers these pathophysiological states remains unknown. Here, we identify an acute microglial metabolic response that rapidly translates intake of high-fat diet (HFD) to a surprisingly beneficial effect on metabolism and spatial/learning memory. High-fat intake rapidly increases palmitate levels in cerebrospinal fluid and triggers a wave of microglial metabolic activation characterized by mitochondrial membrane activation and fission as well as metabolic skewing toward aerobic glycolysis. These effects are detectable throughout the brain and can be detected within as little as 12 hr of HFD exposure. In vivo, microglial ablation and conditional DRP1 deletion show that the microglial metabolic response is necessary for the acute effects of HFD. 13C-tracing experiments reveal that in addition to processing via β-oxidation, microglia shunt a substantial fraction of palmitate toward anaplerosis and re-release of bioenergetic carbons into the extracellular milieu in the form of lactate, glutamate, succinate, and intriguingly, the neuroprotective metabolite itaconate. Together, these data identify microglia as a critical nutrient regulatory node in the brain, metabolizing away harmful fatty acids and liberating the same carbons as alternate bioenergetic and protective substrates for surrounding cells. The data identify a surprisingly beneficial effect of short-term HFD on learning and memory.