A connectivity-constrained computational account of topographic organization in primate high-level visual cortex

Abstract

Inferotemporal cortex (IT) in humans and other primates is topographically organized, with multiple domain-selective areas and other general patterns of functional organization. What factors underlie this organization, and what can this neural arrangement tell us about the mechanisms of high level vision? Here, we present an account of topographic organization involving a computational model with two components: 1) a feature-extracting encoder model of early visual processes, followed by 2) a model of high-level hierarchical visual processing in IT subject to specific biological constraints. In particular, minimizing the wiring cost on spatially organized feedforward and lateral connections within IT, combined with constraining the feedforward processing to be strictly excitatory, results in a hierarchical, topographic organization. This organization replicates a number of key properties of primate IT cortex, including the presence of domain-selective spatial clusters preferentially involved in the representation of faces, objects, and scenes, within-domain topographic organization such as animacy and indoor/outdoor distinctions, and generic spatial organization whereby the response correlation of pairs of units falls off with their distance. The model supports a view in which both domain-specific and domain-general topographic organization arise in the visual system from an optimization process that maximizes behavioral performance while minimizing wiring costs.

Development of object recognition

“…This finding suggests that the mechanisms supporting object recognition in infancy may be different, and more distributed, than those in adults. Indeed, comparisons between infants and ANN models of the ventral visual pathway show that young infants' category representations (4-6 months) are better explained by early layers of the models, which represent simple visual features (Kiat et al, 2021;Xie et al, 2021), than by later layers which are predictive of adult category representations (Blauch, Behrmann, & Plaut, 2022;Rajalingham et al, 2018;Yamins et al, 2014). However, with increased age, infants' category representations become increasingly better described by higher-level layers of the models (Kiat et al, 2021), with the category representation of 18-month-olds being predicted by the multivariate response of adult OTC (Spriet, Abassi, Hochmann, & Papeo, 2021).…”

“…Although visual experience is necessary for the development of mature object recognition abilities, innate constraints and biases guide how those abilities develop (Hasson et al, 2002;Kamps et al, 2020). For instance, category representations in OTC are guided by anatomical constraints such as retinotopy and connectivity (Arcaro & Livingstone, 2021;Blauch et al, 2022). Category representations for items that are typically foveated, such as manipulable objects, faces, and words, develop along OTC regions with preferential connectivity to foveal cortex in V1 (Gomez et al, 2019;Hasson et al, 2002;Kamps et al, 2020;Levy, Hasson, Avidan, Hendler, & Malach, 2001).…”

GLMsingle: a toolbox for improving single-trial fMRI response estimates

“…These exciting “condition-rich” datasets are large enough to propel the development of computational models of how humans process complex naturalistic stimuli. For example, resources such as the Natural Scenes Dataset (NSD,Allen et al, 2022), BOLD5000 (Chang et al, 2019), and THINGS (Hebart et al, 2019) may be useful for advancing our ability to characterize the tuning (Bao et al, 2020;Li and Bonner, 2021;Long et al, 2018;Kriegeskorte and Wei, 2021; Popham et al, 2021), topography (Blauch et al, 2021; Doshi and Konkle, 2021; Zhang et al, 2021; Lee et al, 2020), and computations (Yamins et al, 2014; DiCarlo et al, 2012; Freeman et al, 2013; Marques et al, 2021; Horikawa and Kamitani, 2017) performed in visual cortex.…”

Modeling Category-Selective Cortical Regions with Topographic Variational Autoencoders

“…Similarly, the principle of 'wiring length minimization' [36,18] can be placed in this category, positing that evolutionary pressure has encouraged the brain to reduce the cumulative length of neural connections in order to reduce the costs associated with the volume, building, maintenance, and use of such connections. Computational models which attempt to integrate such wiring length constraints [38,62,10] have recently have been observed to yield localized category selectivity such as 'face patches' similar to those of macaque monkeys. A second hypothesis for the emergence of category specialization, which has recently gained increasing empirical support, derives its explanatory power from information theory.…”