Taicheng Huang

A stochastic world model on gravity for stability inference

The paper investigates how humans infer object stability using a stochastic world model of gravity, finding that gravity's direction is represented as a Gaussian distribution, which explains perceptual illusions and supports efficient stability inference.

Real-world size of objects serves as an axis of object space

This study shows that real-world object size is an axis in object space, supported by both computational models and human fMRI experiments.

Semantic relatedness emerges in deep convolutional neural networks designed for object recognition

This study shows that deep convolutional neural networks develop semantic relatedness among objects during training, similar to human WordNet structures, without explicit top-down guidance.

DNNBrain: a unifying toolbox for mapping deep neural networks and brains

DNNBrain provides a Python toolbox for exploring internal representations in deep neural networks and the brain, enabling cross-disciplinary research between deep learning and cognitive neuroscience.