Reproduce “Continuous-depth Deep Gaussian Processes” 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
Ying Li, Sanyou Wu, Zi Yang, Qiaochu Xu, Yuhao Liu, Zhidi Lin, Michael Minyi Zhang, Petar Djuric and 1 more
FlowDGP models regression by evolving latent representations continuously through an SDE with a GP-based drift prior. In toy regression, continuous-depth models remain more stable across depths than plain DGP, while residual DGP reduces degradation. On heteroscedastic regression, FlowDGP's test NLL is 1.248 ± 0.008 versus 3.527 ± 1.601 for neural SDE, a comparison that does not isolate the prior's effect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.