Jenelle Feather

Sounds synthesized to decorrelate model predictions across human auditory cortex show an adversarially robust CochResNet50 predicts fMRI responses far better than its standard counterpart, a gap hidden with natural sounds.

The paper introduces a Bayesian adaptive method to estimate individual sensory representations by optimizing stimulus distortions based on perceptual judgments.

This study evaluates how well deep neural networks align with human perception in categorizing natural sounds by testing model metamers across different architectures and training tasks.

The paper evaluates how well deep neural networks model auditory brain responses and finds that some models better predict brain activity than others, with task-specific effects on prediction quality.

The paper generates model metamers to compare invariances in neural networks and humans, finding discrepancies and proposing a new benchmark for model assessment.

The paper uses intracranial recordings to show a neural population in human auditory cortex selectively responsive to musical song.