Reproduce “Identifying Latent Neural Dynamics with Recognition-Parametrised Gaussian Process Dynamical Systems” 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

Arielle Rosinski, Qianqian Feng, Lior Fox, Maneesh Sahani

RP-GPdyn combines Gaussian-process dynamics with recognition-parametrised links to infer compact latent states without a parametric neural-observation generator. It exposes fixed points and local transition structure. In motor-cortex reaching data, neural-only states retain behavioral information on 10 held-out conditions. Adding behavior during training raises pooled velocity R² from 0.702 to 0.814, demonstrating cross-condition decoding rather than causal identification. The model assumes autonomous dynamics.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.