ML Health Monitor: Taking the Pulse of Machine Learning Algorithms in Production

摘要

Bringing the research advances in Machine Learning (ML) to production is necessary for businesses to gainvalue from ML. A key challenge of production ML is the monitoring and management of real-time predictionquality. This is complicated by the variability of live production data, the absence of real-time labels and thenon-determinism posed by ML techniques themselves. We dene ML Health as the real time assessment of MLprediction quality and present an approach to monitoring and improving ML Health. Specifically, a completesolution to monitor and manage ML Health within a realistic full production ML lifecycle. We describe a numberof ML Health techniques and assess their effcacy via publicly available datasets. Our solution handles productionrealities such as scale, heterogeneity and distributed runtimes. We present what we believe is the first solutionto production ML Health explored at both an empirical and complete system implementation level.