Nonlinearity, memory, and phase transitions in animal learning
Year: 2009
Type: article
Abstract: Event Abstract Back to Event Nonlinearity, memory, and phase transitions in animal learning Abstracting from physiological details, we present a theory that suggests an explanation behind critical periods and other phenomena in animal learning as a natural consequence of learning dynamics under a small and realistic set of assumptions. Namely, we describe animal learning as a Bayesian inference problem, where the subject tries to estimate a certain real-valued variable describing the state of the world using sensory data in the context of some a priori assumptions. Further, we assume that, in this process, the following holds. (a) The animal has no access to the entire a posteriori distribution of the estimated quantity, and its estimate is given by randomly sampling the distribution, rather than calculating the posterior expectation. (b) The penalty for learning incorrectly (the loss function) is not Gaussian, but saturates at large mismatch between the state of the world and the estimate, so that finding the mode of the posterior is not a liner process. Finally, (c) the subject’s a priori assumptions about the temporal dynamics of the estimated variable are that is varies continuously, but the dynamics has multiple temporal scales, so that long-term, multi-scale memory is essential to represent the temporal structure of the sensory data. We notice that typical protocols in animal learning experiments often contradict the latter assumption, and, instead, the subject's sensory experience may involve long periods of stationarity punctuated by rapid changes designed to evaluate the response to them. We show that the three assumptions above and this mismatched structure of experiments are sufficient to explain a variety of animal learning phenomena, capturing similarities and differences between different experiments, such as those on critical periods in auditory-visual maps alignment in barn owls and matching experiments in rats. For example, dependence of the typical learning time on the dynamics of changes in the prior experiences follows naturally from the theory. Similarly, critical periods will emerge as phase transitions between local maxima of the posterior distribution, but this requires a saturating loss function, and, therefore, explains their absence in the rat matching experiments. Further, abrupt, single trial learning can be predicted, and the tendency of subjects to reverse to the status quo following a transient learning experience is also easy to explain. While currently available experiments are too crude to verify or reject the theory, we suggest new, realistic experiments that can be used for this goal. To this end, we propose experimental protocols that should introduce critical periods and reversals to status quo into experimental systems where they have not been observed before, or remove these phenomena from systems where they are known to exist. We emphasize that the theory developed here is not a physiological theory, able to explain mechanisms underlying the observed learning phenomena. Instead it should be viewed as a learning-theoretic explanation for why such seemingly suboptimal learning mechanisms and behaviors are expected to emerge given the properties of the surrounding world. Conference: Computational and systems neuroscience 2009, Salt Lake City, UT, United States, 26 Feb - 3 Mar, 2009. Presentation Type: Poster Presentation Topic: Poster Presentations Citation: (2009). Nonlinearity, memory, and phase transitions in animal learning. Front. Syst. Neurosci. Conference Abstract: Computational and systems neuroscience 2009. doi: 10.3389/conf.neuro.06.2009.03.183 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 03 Feb 2009; Published Online: 03 Feb 2009. Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Google Google Scholar PubMed Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.
Author: Nemenman Ilya
Language: English
Topic: Memory and Neural Mechanisms
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W51250678