Reproduce “Spliced Binned-Pareto Distribution for Robust Modeling of Heavy-tailed Time Series” 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

Elena Ehrlich, Laurent Callot, François-Xavier Aubet

SBP forecasts time-varying distributions by combining a robust binned model of the central range with Generalized Pareto models for both tails. A neural network predicts bin probabilities and tail parameters from recent observations, while quantile thresholds adapt to the predicted distribution. On five-minute Twitter mentions of a stock ticker, SBP achieved tail-calibration MAE of 4.24 × 10⁻³ versus 8.77 × 10⁻³ for TCN-SPOT, but had higher error on synthetic data.

Summary by alphaXiv.

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