Reproduce “TRACE: A Generalizable Drift Detector for Streaming Data-Driven Optimization” 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
Yuan-Ting Zhong, Ting Huang, Xiaolin Xiao, Yue-Jiao Gong
TRACE detects drift in streaming optimization by summarizing surrogate prediction errors over sliding windows and comparing recent patterns with a summary of the current environment. Its global and context-guided attention classify drift. Integrated into DASE as TRACE-EA, it reports lower E_DT on DBG F1 D4 than DASE: 2.59e+01 ± 9.10e-01 versus 5.23e+01 ± 1.08e-01; fixed windows can delay detection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.