Variation in the geometry of concept manifolds across human visual cortex

Area of Science:

• Neuroscience
• Computer Vision
• Machine Learning

Background:

• Accurate visual concept classification in the brain relies on the geometry of concept manifolds.
• Understanding how this geometry differs between brain regions and deep neural networks (DNNs) is crucial.

Purpose of the Study:

• To investigate the geometric properties of concept manifolds in human visual cortex and DNNs.
• To determine how these properties influence few-shot linear classification accuracy.
• To compare the geometric strategies employed by the brain and DNNs.

Main Methods:

• Utilized a large fMRI dataset to analyze brain activity patterns.
• Estimated geometric properties of concept manifolds, including geometric Signal and Dimensionality.
• Compared these properties across different visual cortex regions and deep neural network layers.

Main Results:

• Classification accuracy in human visual cortex is primarily driven by the distance between manifold centers (geometric Signal).
• Classification accuracy in early and mid DNN layers is driven by the effective number of manifold dimensions (Dimensionality).
• Geometric Signal and Dimensionality are strongly negatively correlated across brain regions and DNN layers.

Conclusions:

• Human visual cortex and DNNs employ distinct geometric strategies for concept classification.
• Despite differences, both systems operate under a shared constraint linking Signal and Dimensionality.
• This research offers insights into the computational principles of visual processing in biological and artificial systems.