Micha Heilbron

The paper shows that representational magnitude predicts memorability across image and word datasets, suggesting it is a general property of distributed representations.

The paper uses fMRI and deep generative models to show spatial prediction effects across visual cortex during natural vision, revealing distinct prediction regimes in central vs. peripheral vision.

This study compares generative and discriminative recurrence in neural networks, finding they use opposing strategies for achieving robustness in vision.

The paper investigates how expectations, attention, and NMDA receptor blockade modulate sensory processing stages using EEG decoding in a visual stimulus experiment.

The paper uses deep generative modeling to show that mouse visual cortex responses are shaped by spatial predictability of natural image patches, with higher-level features driving prediction effects.

This paper shows that pre-onset brain encoding in language processing can be explained by stimulus dependencies rather than neural prediction.

The paper investigates how perceptual states are represented in early visual areas during bistable motion perception, finding stronger top-down modulation in V1 than in higher-order areas.

This study investigates how neural activity during naturalistic music listening reflects melodic expectations based on short-range musical contexts and long-term statistical learning.

This study uses brain recordings and a deep neural network to show that predictions occur at multiple levels during language comprehension.

The paper examines how prediction and preview influence reading times and skipping in natural reading, finding they do not affect skipping but significantly impact reading times.

This study shows that confidence reports can distinguish hierarchical from non-hierarchical learning models under uncertainty.