Thomas Naselaris

Variation in the geometry of concept manifolds acrosshuman visual cortex

The paper examines how geometric properties of concept manifolds in human visual cortex and DNNs affect classification accuracy, finding differences in the key geometric factors driving performance.

A transformation from vision to imagery in the human brain

The paper introduces an imagery transformation framework to compare visual and mental imagery brain activity, finding that imagery patterns are transformed versions of visual patterns with fewer active dimensions.

Disentangling signal and noise in neural responses through generative modeling

This paper introduces GSN, a method to disentangle signal and noise in neural responses, showing it improves dimensionality estimates in fMRI data.

Vicarious Somatotopic Maps Tile Visual Cortex

The paper maps somatosensory and visual field tuning in the brain, showing aligned somatotopic body maps in the visual cortex during movie watching.

Stacked regressions and structured variance partitioning for interpretable brain maps

This paper introduces stacking and structured variance partitioning methods to improve interpretability of brain maps by handling correlated stimulus features.

Temporal asymmetry of neural representations predicts memory decisions

The paper shows that memory decisions are predicted by the difference in neural similarity between past and future events, with strongest effects in recent memory and weak true memories.

Re-expression of CA1 and entorhinal activity patterns preserves temporal context memory at long timescales

The paper shows that re-expression of CA1 and entorhinal activity patterns predicts temporal context memory over long timescales in humans.

Color-biased regions in the ventral visual pathway are food-selective

The paper analyzes fMRI data to show that color-biased regions in the ventral visual pathway are food-selective and respond to color saturation and other image features.

Natural language supervision with a large and diverse dataset builds better models of human high-level visual cortex

The paper shows that using natural language supervision with large, diverse datasets improves models' ability to predict human high-level visual cortex responses compared to image-only or text-only models.

Brain-optimized neural networks learn non-hierarchical models of representation in human visual cortex

This study trains neural networks to predict human visual cortex activity and finds they can learn non-hierarchical representations, challenging the idea that visual areas function in a strict hierarchical sequence.

A massive 7T fMRI dataset to bridge cognitive and computational neuroscience

The paper introduces a large-scale 7T fMRI dataset of natural scenes to advance cognitive and computational neuroscience research.

Voxel-to-voxel predictive models reveal unexpected structure in unexplained variance

The paper uses voxel-to-voxel models to analyze unexplained variance in BOLD responses and finds it is structured within and across visual areas but not across brains.

Human brain activity during mental imagery exhibits signatures of inference in a hierarchical generative model

The paper investigates brain activity during mental imagery and finds it mirrors visual processing through an echo effect in hierarchical encoding models.

Generative Adversarial Networks Conditioned on Brain Activity Reconstruct Seen Images

This paper uses GANs conditioned on brain activity to reconstruct images from fMRI data, showing they can recover basic stimulus outlines despite noise and limited data.

The feature-weighted receptive field: an interpretable encoding model for complex feature spaces.

The paper introduces the feature-weighted receptive field (fwRF) model, which balances expressiveness, interpretability, and scalability for encoding complex feature spaces, and demonstrates its application to visual neuroimaging data.