Michelle R. Greene

This paper develops a brain-guided CNN to study how task-specific features are processed in scene categorization, finding that task-relevant features emerge at different times during processing.

The paper introduces a method to visualize the spatiotemporal code of visual information using EEG and state-space theory, revealing dynamic neural transformations across electrodes and images.

This study uses whitening transformations to disentangle visual and conceptual features' contributions to scene categorization, finding high-level features drive behavior while low-level features initiate early vERP responses.

The paper investigates the information carried by steady-state visual evoked potentials (SSVEPs) by mapping high-dimensional responses to a 2D space and introducing a method to measure mutual information.

This study compares sequential visual representations in deep CNNs and human ERPs, finding temporal alignment between early CNN layers and early ERP activity, and conceptual CNN layers with later ERP activity.

The paper assesses how functional and deep neural network features contribute to scene representation in human brain and behavior, finding distinct roles for each in explaining similarity in behavior and neural responses.