James J. DiCarlo

The paper shows predictive coding models replicate human perception and neuronal responses to motion-defined shapes in videos.

The paper models plasticity in macaque inferior temporal cortex after object training using ANN simulations, showing gradient descent approximates observed neural changes.

The paper shows that primate amygdala neurons have visual selectivity linked to valence without conditioning, using image-computable models to predict this relationship.

The paper shows that some neurons in the human amygdala and hippocampus encode face identities based on shared visual features rather than invariant semantic concepts.

The paper tests if topographic deep ANN models of the primate ventral stream can predict perceptual effects of direct IT cortical interventions using perturbation modules and benchmarking against experimental results.

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.

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 improves AI models by aligning their representations with macaque brain data, enhancing behavioral alignment and adversarial robustness.

The paper develops a chronically implantable LED array for optogenetic perturbation in non-human primates and demonstrates its effectiveness in inducing visual deficits.

The paper predicts human behavioral changes in object recognition from unsupervised neural plasticity in IT cortex using a computational model.

The paper builds and evaluates hierarchical neural network models to approximate primate V1 neurons and shows their alignment with human object recognition behavior.

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.

This study shows that ventral prefrontal cortex (vPFC) is critical for recurrent processing in macaque IT cortex, which is necessary for robust visual object recognition of late-solved images.

The paper investigates neuronal representations of letters and letter strings in macaque monkeys and shows that IT cortex can support orthographic processing tasks similar to baboon behavior.

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.

A fixed-weight, biologically parameterised V1 front-end gives standard image classifiers adversarial and corruption robustness at no extra training cost; its simulated neuronal noise drives most of the adversarial gain.

This paper reduces the number of supervised synaptic updates needed to model the primate ventral visual stream, achieving 80% of the adult match with significantly fewer updates.

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.

The paper finds that model performance on monkey IT data predicts performance on human IT data, outperforming ImageNet-based predictors.

The paper uses deep image synthesis to control primate V4 neural activity, achieving population-level control beyond natural levels.

The paper shows that recurrent circuits in the primate ventral stream are critical for object recognition, as challenge images require longer processing and are poorly predicted by feedforward models.

This study inactivates specific IT regions in primates to show that different sub-regions contribute uniquely to object recognition, with deficits correlating to neuronal discriminability and anatomical proximity.

The paper shows that neural dynamics in the ventral visual stream involve hierarchical error signals between cortical stages, captured by recurrent models rather than standard feedforward circuits.

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.

This paper compares human, monkey, and deep neural network performance in object recognition at the image level, finding that current models fail to predict primate behavior.

This paper introduces a novel microendoscope using ultrathin optical fibers for deep brain imaging with micron-scale resolution and improved speed in vivo.

This paper builds a deep convolutional neural network optimized for object recognition and shows its hierarchical representations align with human ventral visual stream areas from V1 to LOC.