Michael H. Herzog

This study shows that increased inter-trial latency variability and reduced neural responses both contribute to diminished visual EEG N1 amplitudes in schizophrenia.

This study shows humans persistently explore stochastic environments despite suboptimal strategies, driven by novelty rather than optimal information gain.

This study investigates reinforcement learning in older adults using a high-demand navigation task, finding no significant age-related differences in key behavioral and model parameters.

The paper uses EEG in the SQM paradigm to identify two distinct neural stages of visual processing: an early unconscious stage and a later conscious stage.

This study determines the minimum number of phosphenes needed for object recognition in prosthetic vision, finding that 65 phosphenes suffice, with segments requiring fewer fragments.

The paper re-evaluates the correlation between visual illusion susceptibility and neural measures, finding no significant correlations at the perceptual or neural levels.

This study investigates predictive coding abnormalities in schizophrenia using EEG data, finding altered top-down and bottom-up alpha-band traveling waves.

This study introduces a deep sequential decision-making paradigm to explain human behavior by incorporating novelty and surprise, showing that model-free actions dominate decisions despite the availability of a world-model.

This study examines the neural correlates of visual backward masking in bipolar disorder patients, finding reduced global field power amplitudes similar to schizophrenia.

The paper shows that Capsule Neural Networks (CapsNets) with recurrent grouping and segmentation outperform feedforward and standard recurrent CNNs in modeling human global shape processing using visual crowding.

This study uses visual crowding to show that feedforward CNNs lack human-like global shape processing due to architectural limitations.

This paper introduces a synaptic plasticity model that enables fast learning without catastrophic forgetting by combining early weight decay and selective consolidation.

The paper examines the unsupervised bias hypothesis for failure to learn during roving tasks and finds it insufficient, proposing an alternative explanation involving certainty-normalized reward signaling.