Computational implications of a normalized value representation in decision circuits

Abstract

Event Abstract Back to Event Computational implications of a normalized value representation in decision circuits Value information is a critical component of decision-making, and a growing number of studies report value-related signals in decision-related neural circuits. Decision theorists have long hypothesized that choices are guided by a relative reward representation, but the specific form in which the brain encodes value remains unclear. Using two- and three-target choice tasks, our recent work suggests that visuomotor neurons in the monkey lateral intraparietal area (LIP) encode a relative measure of saccade value, normalized across available options: activity is negatively correlated to target values outside the response field while positively correlated to target value in the response field. Interestingly, this value normalization is well-described by a divisive normalization model that also characterizes nonlinear phenomena such as gain control and cross-orientation suppression in visual cortex: FR = K*[B+V] / [S+Vt] where the firing rate of an LIP neuron (FR) is a function of the value associated with a saccade to the target in the response field (V), the total value of all available options (Vt), and three additional parameters (K,B,S). How might such a value-coding scheme affect choice behavior? Here, we explore the computational implications of divisive normalization for value-based decision-making. Specifically, we examine how relative reward encoding interacts with noisy population estimates of value and the resulting effects on choice. The motivating intuition is straightforward: as the total value of available options increases, the separation between the distributions of firing rates representing two differently valued options decreases; if variance doesn’t decrease appropriately, the options will be increasingly difficult to distinguish. To examine this computationally, we examine the choice between two differently valued options as additional lower valued options are added to the choice set. We simulate the decision process as the selection of the highest valued option, with the value of each option drawn from a distribution of firing rates around the option-specific mean firing rate. The mean rate for each option is determined using the value-based divisive normalization equation, with parameters fit to the observed monkey physiological data. Noisy population coding is modeled by including two general sources of noise: 1) a mean-rate dependent noise term, which models the considerable intrinsic response variability observed in cortical neurons (response variance typically ~1-1.5 times the mean response), and 2) a mean-rate independent additive noise term. We find that the presence of either form of noise leads to increasing errors as the choice set size increases, and characterize this phenomenon as a function of the number and value of additional alternatives as well as different degrees of noise. We conclude that a relative representation of reward, mediated by divisive normalization, can lead to suboptimal choice behavior. These results suggest a computational explanation for why and how choice behavior responds to multiple options and changing choice sets, and highlight the importance of understanding both the specific mechanism of cortical representation and the nature of the population code. 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). Computational implications of a normalized value representation in decision circuits. Front. Syst. Neurosci. Conference Abstract: Computational and systems neuroscience 2009. doi: 10.3389/conf.neuro.06.2009.03.218 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.