Hierarchical Models of the Visual System

doi:10.1007/978-1-4614-7320-6_345-2

Paper

Hierarchical Models of the Visual System

Published 2020 · Matthew Ricci, Thomas Serre

UNKNOWN SJR score

1

Citations

0

Influential Citations

Full text

Semantic Scholar

Abstract

Hierarchical models of the visual system are neural networks with a layered topology. The receptive fields of units (i.e., the region of visual space to which units respond) at one level of the hierarchy are constructed by combining inputs from units at a lower level. After a few processing stages, small receptive fields tuned to simple stimuli get combined to form larger receptive fields tuned to more complex stimuli. Such an anatomical and functional hierarchical architecture is a hallmark of the organization of the visual system. In feedforward networks, information flows in a bottom-up fashion – from lower to higher processing stages. In feedback networks, information is able to dynamically reenter processing stages via recurrent connections. Feedback connections can be broadly divided between horizontal or lateral connections within processing stages and top-down connections from higher onto lower processing stages. Since the pioneering work of Hubel andWiesel (1962), a variety of hierarchical models have been proposed, from relatively small-scale models of the primary visual cortex to very large-scale (system-level) models of object and action recognition, which account for visual processing in entire visual streams. The term “model of the visual system” is generally reserved for architectures that are constrained in some way by the anatomy and the physiology of the visual system (with various degrees of realism). Deep convolutional networks are architecturally similar neural networks that have led to impressive results in a wide range of engineering disciplines, from computer vision to natural language processing, and artificial intelligence more broadly.

Study Snapshot

Key takeawayHierarchical models of the visual system are neural networks with a layered topology, reflecting the anatomical and functional hierarchy of the visual system.

PopulationOlder adults (50-71 years)

Sample size24

MethodsObservational

OutcomesBody Mass Index projections

ResultsSocial networks mitigate obesity in older groups.

Sign up to use Study Snapshot

Consensus is limited without an account. Create an account or sign in to get more searches and use the Study Snapshot.

Create an account

Paper

Hierarchical Models of the Visual System

Sign up to use Study Snapshot

References

For Further Reading

Few people in West Africa use climate research results and few have access to seasonal forecasts, despite a majority willingness to do so.

Unsupervised Learning

Unsupervised learning focuses on learning to represent input patterns based on their statistical structure, rather than explicitly target outputs or environmental evaluations.

Ducklings imprint on the relational concept of “same or different”

Ducklings can learn relational concepts, even without reinforced training, revealing their ability to process abstract properties and apply them to novel stimuli.

Toward an Integration of Deep Learning and Neuroscience

The brain optimizes diverse cost functions within a pre-structured architecture, similar to machine learning, making learning data-efficient and targeted to the organism's needs.

Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence

Deep neural networks effectively capture the stages of human visual processing, providing an algorithmically informed view on the spatio-temporal dynamics of visual object recognition in the human visual brain.

Citations

EAGLE: Eyegaze-Assisted Guidance and Learning Evaluation for Lifeloging Retrieval

EAGLE optimizes lifelog information retrieval by incorporating user eye movements, supporting both novice and experienced users, and minimizing effort in the search process.

Video-Language Models as Flexible Social and Physical Reasoners

This paper presents a novel video-language architecture that incorporates spatiotemporal embeddings and hidden states, enabling flexible reasoning in both social and physical contexts.

Visual shape discrimination in goldfish, modelled with the neural circuitry of optic tectum and torus longitudinalis

Goldfish can discriminate shapes visually using the optic tectum and torus longitudinalis, with a model that biases attention to edges and mimics their natural behavior.

Eye-specific detection and a multi-eye integration model of biological motion perception

Jumping spiders can discriminate biological motion from random displays using only their anterior-lateral eyes, suggesting a multi-eye integration system for visual processing.

Neural correlates of visual object recognition in rats

Rats excel in visual object recognition, with species-specific differences in neural circuits supporting this capacity.

Septotemporal variations in hippocampal value and outcome processing.

The dorsal hippocampus represents external value landscape, while the ventral hippocampus preferentially represents behaviorally relevant, salient features of experienced events.

Computational modeling of color perception with biologically plausible spiking neural networks

Our neuromorphic Spiking Neural Network-driven model accurately reconstructs colorful images from retinal inputs, allowing for critical evaluation of visual phenomena in a biologically plausible computational framework.

Neural similarity between overlapping events at learning differentially affects reinstatement across the cortex

Differentiation and integration processes in the brain's cortex have divergent effects on recall of highly similar naturalistic events.