H. Steven Scholte

The paper shows that deep neural networks align more closely with human brain responses to texture-synthesized images than to natural or object-only images, suggesting shared texture statistics rather than object recognition drive this alignment.

This study shows brain alignment in deep neural networks is present early and independent of object classification, suggesting it reflects environmental structure rather than classification strategy.

The paper uses image-grounded encoding models to show that object processing in the human brain has a delayed temporal profile compared to scene processing in natural visual environments.

The paper introduces object-zoomed training for CNNs to improve shape bias, aligning CNN performance with human-like object recognition strategies.

This study improves DNN encoding models of human visual processing by aligning their spatial sampling with human retinal ganglion cell density, revealing distinct temporal profiles for central and peripheral visual information.

This paper shows that increasing viewpoint diversity in training improves generalization to novel views, aligning models more closely with human object recognition strategies.

The paper introduces a neuroimaging dataset of autistic and non-autistic adults with extended functional scans to study individual differences in brain function.

The paper investigates why deep convolutional neural networks (DCNNs) with higher object recognition accuracy perform worse at predicting EEG responses, finding that EEG signals reflect scene segmentation rather than object-specific representations.

This study shows that both human visual cortex and deep convolutional neural networks prioritize object background over category in object recognition tasks.

This paper develops deep learning models to study dynamic object recognition, finding that sequential processing with lateral recurrence matches human performance and improves recognition with adaptation.

The paper presents a computational model showing how arousal state modulates task performance by altering sensory processing hierarchy, explaining the Yerkes-Dodson effect.

This study uses spiking deep neural networks to compare attention mechanisms, finding that modulating outgoing connections is more effective than precision or input gain in enhancing visual search performance.

This paper tests if deep residual networks (ResNets) can capture recurrent processing in the human brain using EEG data from a visual masking task, finding deeper networks better explain brain activity.

This study shows that recurrent processing is critical for object recognition in complex scenes, as evidenced by behavioral, EEG, and deep learning results.

The paper introduces the Amsterdam Open MRI Collection, a multimodal MRI dataset with structural and functional data from healthy participants, available for individual difference analyses.

The paper shows that choosing negative stimuli activates reward circuitry similar to extrinsic incentives and curiosity, contrasting with positive stimuli choices.

This paper shows that deeper convolutional neural networks better distinguish objects from backgrounds in scenes, suggesting that scene segmentation can emerge from network depth without explicit mechanisms.

The paper shows that feedback activity in the brain increases during object recognition in complex natural scenes, improving performance, while feed-forward processing suffices for simpler scenes.

The paper analyzes rare coding variants in smoking and drinking behavior using exome chip data, finding that they contribute to phenotypic variation but require larger samples for individual variant associations.

This paper evaluates counterbalancing and confound regression for controlling confounds in decoding analyses and finds that foldwise confound regression is the only effective method.

The paper uses multi-task deep neural networks to analyze visual pathway organization, finding that unrelated tasks lead to less feature sharing in higher layers.

This study examines how the depth of convolutional neural networks (CNNs) correlates with EEG responses during visual object recognition, showing that deeper CNNs improve predictions after 150 ms.