Science Advances · 2022 · Journal article

A data-based large-scale model for primary visual cortex enables brain-like robust and versatile visual processing

Guozhang Chen, Franz Scherr, Wolfgang Maass

A comprehensive data-based model of primary visual cortex achieves versatile, noise-robust visual processing with sparse, energy-efficient neural coding.

The scientific question

Can an anatomically realistic primary visual cortex model support versatile visual processing?

Convolutional neural networks lack biological circuit structure, consume substantial energy, and remain fragile to sensory noise compared with mammalian vision.

Data-based model of mouse primary visual cortex.
Figure 1 from the original paper ↗ · CC BY 4.0

The core idea

Multi-Task Evaluation of Bio-Realistic V1

We evaluate a large-scale spiking model of area V1 incorporating detailed anatomical connectivity and neurophysiological cell properties across diverse spatial and temporal visual tasks.

What we found

Brain-Like Versatility and Noise Robustness

A single data-driven V1 circuit solves diverse visual tasks, integrates temporally dispersed inputs, reproduces cortical coding properties, and exhibits strong resilience to perturbation.

Continue the question

Related reading

Explore our research

Cite this work

Guozhang Chen, Franz Scherr, Wolfgang Maass. A data-based large-scale model for primary visual cortex enables brain-like robust and versatile visual processing. Science Advances, 2022. doi:10.1126/sciadv.abq7592.