Tupl
ML Performance Drift Detection and Correction with Time Series and Active Learning
Machine Learning (ML) models drift and show degradation in performance as real-world data change over time. Therefore, these models need to be retrained with new data that reflect the new reality of the process. The student team implemented an ML Drift Detection module to automate detecting and measure drift, and an Active Learning module to retrain the model using new data in an effective way. The solution is implemented as a microservices REST backend with an interactive UI for the user to input necessary parameters and to show the vital parameters of the ML model.
Faculty Adviser
Payman Arabshahi,
Associate Professor, UW ECE,
Electrical & Computer Engineering
Students
Alaa Sleek
Chun Yen
Matthew Varghese
Sreejith S