TRACE: A Generalizable Drift Detector for Streaming Data-Driven Optimization

TL;DR

TRACE is a novel drift detection method that uses attention-based sequence learning and tokenization to accurately identify distributional changes in streaming data, improving adaptability and robustness in dynamic optimization environments.

Contribution

We introduce TRACE, a transferable, attention-based drift detector that generalizes across datasets and integrates seamlessly with streaming optimizers for adaptive data-driven optimization.

Findings

Outperforms existing drift detectors on diverse benchmarks.

Demonstrates high transferability to unseen datasets.

Enhances adaptive optimization under concept drift.

Abstract

Many optimization tasks involve streaming data with unknown concept drifts, posing a significant challenge as Streaming Data-Driven Optimization (SDDO). Existing methods, while leveraging surrogate model approximation and historical knowledge transfer, are often under restrictive assumptions such as fixed drift intervals and fully environmental observability, limiting their adaptability to diverse dynamic environments. We propose TRACE, a TRAnsferable C}oncept-drift Estimator that effectively detects distributional changes in streaming data with varying time scales. TRACE leverages a principled tokenization strategy to extract statistical features from data streams and models drift patterns using attention-based sequence learning, enabling accurate detection on unseen datasets and highlighting the transferability of learned drift patterns. Further, we showcase TRACE's plug-and-play nature by integrating it into a streaming optimizer, facilitating adaptive optimization under unknown drifts. Comprehensive experimental results on diverse benchmarks demonstrate the superior generalization, robustness, and effectiveness of our approach in SDDO scenarios.