A Population Response Model of Ensemble Perception

Abstract

Ensemble representations have been considered as one of strategies that the visual system adopts to cope with its limited capacity. Thus, they include various statistical summaries such as mean, variance, and distributional properties and are formed over many stages of visual processing. The current study proposes a population coding model of ensemble perception to provide a theoretical and computational framework for these various facets of ensemble perception. The proposed model consists of a simple feature layer and a pooling layer. We assumed ensemble representations as population responses in the pooling layer and decoded various statistical properties from population responses. Our model successfully predicted averaging performance in orientation, size, color, and motion direction across different tasks. Furthermore, it predicted variance discrimination performance and the priming effects of feature distributions. Finally, it explained the well-known variance and set size effects and has a potential for explaining the adaptation and clustering effects.

Ensemble averaging: What can we learn from skewed feature distributions?

“…One model is the loss function model suggested by Teng et al (2021) for the perception of skewed distributions that suggests a formal mathematical algorithm of evidence sampling and decision making based on this evidence. Another model is a neurally plausible population coding model ( Utochkin et al, 2022 ) with three versions of decoding rules. These two kinds of models are not mutually exclusive but rather represent two different levels of models, in Marr's (1982) classification.…”

“…The pooled population response has a lot of properties that are isomorphous to the feature distribution of a presented set that make it possible to read out various statistical information: not only central tendency (by peak location, although other decoding rules are also possible: Georgopoulos et al, 1986 ; Jazayeri & Movshon, 2006 ; Ma, Beck, Latham, & Pouget, 2006 ), but also variance (the width of the population response distribution) and even the shape of the distribution (e.g., whether this distribution has a single peak or several peaks). Computational simulations based on this model successfully capture ensemble perception patterns across statistical summaries, feature domains, and paradigms reported in the literature ( Utochkin et al, 2022 ).…”

“… Dakin, Mareschal, and Bex (2005) and Webb, Ledgeway, and McGraw (2007 , 2010 ) suggested population code models of motion and orientation integration, although their main focus was on distinguishing between different decoding mechanisms that would better account for behavioral data. In their recent work, Utochkin, Choi, and Chong (2022) , developed a general population coding model of ensemble perception. They shift the main focus to pooling itself as a key part of ensemble encoding process and conceptualize the pooled population response as the neural representation of ensemble.…”

Ensemble Representation of Animacy Is Based on Mid-level Visual Features

“…A general explanation of how we compute summary statistics frequently involves pooling models: information about visual features is pooled from the lower areas of the visual cortex, "mixed up", and averaged in the higher areas (Parkes et al, 2001;Balas et al, 2009;Haberman & Whitney, 2011;Rosenholtz et al, 2012;Whitney & Yamanashi Leib, 2018;Utochkin, Choi, Chong, 2022;Robinson & Brady, 2022). However, some studies (Yamanashi Leib et al, 2016;Whitney & Yamanashi Leib, 2018) suggested that pooling models are insufficient to explain such high-level ensemble properties as mean animacy, assuming a unique mechanism behind this process.…”

A signal-detection account of item-based and ensemble-based visual change detection: A reply to Harrison, McMaster, and Bays

“…Three recent works proposed various quantitative population-like models that account for ensemble representations and the role of relational information in visual working memory. Utochkin, Choi, & Chong (2022) suggest a pooling and population coding model that captures the variety of ensemble representations, including the computation of average and variability, as well as setting boundary conditions for all items being perceived as belonging to one or several clusters. Robinson and Brady (2022) conceptualize the ensemble representation as a pooled familiarity signal produced by individual items in noisy feature-selective channels (Schurgin, Wixted, & Brady, 2020), an idea having a strong reference to population coding.…”