Daniel L. K. Yamins

The paper uses a self-supervised neural network to show that visual streams in the brain may arise from a single computational principle rather than separate behavioral specializations.

The paper identifies a common temporal motif in neural dynamics across the ventral visual stream and shows that current models fail to replicate these dynamics.

This study shows that diverse perceptual representations in visual pathways arise from a single objective of understanding world dynamics, using deep neural network models to predict human neural data.

The paper shows that the perirhinal cortex integrates visuospatial sequences to support object perception, complementing the visual features processed by VTC.

This paper introduces TDANN, a model that predicts functional organization in primate visual cortex by balancing task-general representation and spatial smoothness, showing improved brain-like representations and applications in prosthetic design.

The paper reassesses claims about grid cell emergence in path-integrator circuits, finding consistency with prior theory.

This study develops ConvRNN models to simulate primate ventral visual stream processing, finding that recurrent networks achieve better task performance and efficiency than deeper feedforward networks.

This study uses task-driven neural networks to show that heterogeneous MEC cells are functionally significant and that reward-modulated models better capture MEC response profiles than traditional models.

This paper uses a deep learning framework to model visual discrimination tasks and shows that PRC-intact participants outperform both VVS models and PRC-lesioned behavior.

This paper develops a computational model of mouse visual cortex, finding that self-supervised contrastive learning produces better matches to mouse cortex than supervised methods, suggesting a resource-efficient, general-purpose visual system.

This paper shows that unsupervised neural networks can model primate ventral visual stream responses as well as supervised methods, with brain-like representations from limited data.

Adding a wiring-length penalty to object-recognition training makes model inferior-temporal layers grow face-selective units, face patches, inter-patch connectivity and viewpoint invariance, but only under naturalistic training input.

This study compares CNN layer representations with EEG data to find hierarchical and non-hierarchical correlations in object processing timing.

Among dozens of ImageNet-trained networks scored against macaque V4/IT recordings and human behavior, DenseNet-169, CORnet-S and ResNet-101 rank highest, and brain similarity stops tracking accuracy above 70% top-1.

This study investigates how visual production and recognition of objects share neural representations in visual cortex and how drawing practice enhances perceptual discriminability.

CORnet-S, a four-area recurrent network with areas pre-mapped to V1, V2, V4 and inferior temporal cortex, ranks among top Brain-Score models while having a shorter longest convolutional path than competitors.

The paper investigates how drawing recruits visual feature representations similar to those used in object recognition, showing that learning to draw improves recognition of objects.