bioRxiv · 2023 · Preprint

Data-based large-scale models provide a window into the organization of cortical computations

Guozhang Chen, Franz Scherr, Wolfgang Maass

Training data-based large-scale cortical microcircuit models enables reverse-engineering of how anatomical structure organizes sparse, energy-efficient cortical computation.

The scientific question

How does the anatomical structure of cortical microcircuits organize neural computation?

Although neurobiology has mapped neocortical cell types and connectivity, experimental tools alone cannot fully resolve causal structure-function relationships.

Structure and components of the data-based cortical microcircuit model.
Figure 1 from the preprint ↗ · CC BY-NC-ND 4.0

The core idea

Reverse-Engineering Data-Driven Microcircuits

We train large-scale spiking models integrating detailed anatomical and physiological microcircuit data on natural tasks, then analyze internal dynamics to reverse-engineer their computational organization.

What we found

Sparse Coding and Testable Hypotheses

Trained microcircuits reproduce experimental cortical coding traits absent in standard deep networks, compute with energy-efficient sparse activity, and generate testable hypotheses for neuroscience.

Continue the question

Related reading

Explore our research

Cite this work

Guozhang Chen, Franz Scherr, Wolfgang Maass. Data-based large-scale models provide a window into the organization of cortical computations. bioRxiv, 2023. doi:10.1101/2023.04.28.538662.