Reproduce “A Conformal Prediction Approach to Explore Functional Data” 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

Jing Lei, Alessandro Rinaldo, Larry Wasserman

Jing Lei, Alessandro Rinaldo, and Larry Wasserman develop a distribution-free framework for exploring functional data through conformal prediction. The method produces simultaneous prediction bands and hierarchical cluster trees with finite-sample coverage guarantees, illustrated on neural recordings and phoneme periodograms with empirical coverages of 91.6% and 90.5% at nominal 90% levels.

Key insights

  • Conformal calibration supplies predictive validity without requiring a parametric distribution, while still allowing flexible scores based on mixtures, pseudo-densities, or learned projections.
  • Projection-based bands can represent multimodal structure and may have disconnected pointwise sections, unlike summaries designed around a single central functional region.
  • The resulting cluster tree is a calibrated visualization of high-concentration regions rather than a guarantee that the estimated number or scientific meaning of clusters is correct; its structure depends on the basis, distance, bandwidth, and connectivity threshold.

Summary by alphaXiv.

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