Reproduce “ProbRes: Volatility Learning for Probabilistic Time-Series Forecasting” 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
Tingting Wang, Yunyi Zhang, Benyou Wang
ProbRes turns a point forecaster into a probabilistic forecaster by modeling the conditional mean, changing conditional volatility, and the distribution of normalized residuals. It learns volatility from transformed errors, rescales residuals, and resamples them to generate future paths. On Exchange, DLinear with ProbRes reached CRPS 0.010 versus 0.017 for bootstrap, with overlapping margins; theory is conditional.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.