Reproduce “Modeling Visual Cortex with Layerwise Multiscale Manifold Capacity Maximization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Thomas Yerxa, SueYeon Chung, Eero P. Simoncelli
ST-MMCR trains visual stages with local objectives, scaling spatial pooling and temporal integration to receptive fields rather than propagating errors across the network. Using one evolving video stream, it predicts macaque V1, V2, and V4 responses about as well as adversarially trained AlexNet and better than supervised AlexNet. Its ImageNet-label-trained V2-front-end classifier reaches 0.485 average out-of-distribution top-1 accuracy versus 0.434 for supervised AlexNet.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.