Reproduce “Embedding stochastic dynamics in latent manifolds of recurrent 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
Ramón Nartallo-Kaluarachchi, Renaud Lambiotte, Alain Goriely
The authors train low-rank recurrent neural networks to reproduce specified stochastic dynamics by matching the target drift and diffusion within a learned affine latent subspace. The method uses additive noise and constrains trajectories to that subspace. In demonstrations, 128-neuron networks reproduce planar processes with three, four, and five attractors; these simulations illustrate memory models, not biological memory.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.