Nikolaus Kriegeskorte

This paper compares recurrent convolutional neural networks with feedforward models for object recognition, finding recurrent networks more accurate and robust, especially under occlusion and noise.

The paper compares computational models of face recognition using fMRI data and finds that a face-space coding model with sigmoidal ramp tuning performs better than an image-processing model.

The paper investigates how brain activity links visual stimuli to behaviour, finding that behaviour-related activity emerges rapidly in the ventral visual pathway and cannot be explained by visual features or semantic categories alone.

The paper develops a unified mathematical framework for encoding analysis, pattern-component modeling, and representational similarity analysis, showing they all evaluate the second moment of activity profiles and are complementary tools for understanding neural representations.

The paper compares fixed and mixed representational similarity analysis (RSA) to explain visual representations in primate brain areas using shallow and deep computational models.

The paper compares categorical and feature-based models in explaining IT cortex representations and similarity judgments, finding that shared variance between features and categories best explains IT representations, while similarity judgments are better explained by categories alone.

This study shows that fMRI can detect fine-grained orientation preferences in V1, challenging the global-areal-map hypothesis and demonstrating that local neuronal selectivity contributes to orientation decoding.

This paper shows that temporal pattern drift in fMRI data affects pattern similarity estimates and proposes methods to recover orientation information.

The paper tests the faciotopy hypothesis, finding that the OFA shows evidence of a face-feature map with cortical distances reflecting physical feature distances in faces.

The paper shows that intrinsic cortical dynamics influence representational geometry in human visual cortex, making responses to natural images more similar across brain regions when intrinsic dynamics are shared.

This paper reviews advances in neuroimaging analysis that move beyond traditional averaging by modeling individual representational idiosyncrasies using multivariate pattern analysis and representational dissimilarity matrices.