The representational geometry for naturalistic textures in macaque V1 and V2

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Processing

Background:

• Visual cortical processing traditionally studied via individual neuron selectivity.
• Representational geometry offers a complementary approach to understand population encoding strategies.
• Distinct functional objectives may relate to different representational geometries.

Purpose of the Study:

• Investigate changes in representational geometry across visual processing stages.
• Compare neuronal population responses in primary visual cortex (V1) and area V2.
• Understand texture encoding transformations between V1 and V2.

Main Methods:

• Measured neuronal population responses in macaque V1 and V2.
• Utilized synthetic and naturalistic textures as stimuli.
• Analyzed representational geometry and compared with convolutional neural network models.

Main Results:

• Representational geometry was lower-dimensional in V2 compared to V1.
• V2 population responses showed better alignment and discriminability for different textures.
• Standard convolutional network models did not replicate the observed cortical representational geometry.

Conclusions:

• A significant shift in representational geometry occurs between V1 and V2.
• V2 exhibits a low-dimensional, systematic encoding of textures, enhancing discriminability.
• Comparing representational geometry is crucial for understanding visual processing transformations.