Reproduce “Structured Visual Landscape: Generating Preferred Representations in Biological and Artificial Neural Networks” 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
Yusong Zhao, Edgar Y. Walker, Hao Wang, Mi Lu
Vis-Lens generates images to increase or decrease selected brain-region or network-unit responses. It trains a variational autoencoder with separate latent centers for positive and negative responses, then edits an image’s latent code and uses a frozen diffusion model to generate a stimulus. Across seven NSD ROIs, predicted decreases were stronger, including −0.699 versus +0.186 for IP0; these are simulator predictions, not new human measurements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.