Reproduce “GALACTIC: Global and Local Agnostic Counterfactuals for Time-series Clustering” 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
Christos Fragkathoulas, Eleni Psaroudaki, Themis Palpanas, Evaggelia Pitoura
GALACTIC produces local counterfactuals—minimally changed time series that a surrogate assigns to another cluster—and summarizes recurring transitions across a population. It segments series, estimates segment importance, and restricts gradient edits with source-and-target masks; an MDL objective selects a compact set of edits. On ArrowHead, Greedy GALACTIC-G reaches 75.9% effectiveness, versus 54.6% for adapted GLOBE-CE, but has higher average flipping cost than adapted Glacier-G. Results concern surrogate assignments on benchmark labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.