ML models are increasingly being pushed to mobile devices, for low-latency inference and offline operation. However, once the models are deployed, it is hard for ML operators to track their accuracy, which can degrade unpredictably (e.g., due to data drift). We design the first end-to-end system for continuously monitoring and adapting models on mobile devices without requiring feedback from users. Our key observation is that often model degradation is due to a specific root cause, which may affect a large group of devices. Therefore, once the system detects a consistent degradation across a large number of devices, it employs a root cause analysis to determine the origin of the problem and applies a cause-specific adaptation. We evaluate the system on two computer vision datasets, and show it consistently boosts accuracy compared to existing approaches. On a dataset containing photos collected from driving cars, our system improves the accuracy on average by 15%.

QUETTA, PAKISTAN – The family of a driver who was killed alongside Taliban chief Mullah Akhtar Mansour in a U.S. drone strike in Pakistan has filed a case against U.S. officials, seeking to press murder charges, police said Sunday. Mansour had entered Pakistan from Iran using a false name and fake Pakistani identity documents on May 21, when his car was targeted by a U.S. drone. The driver, who was also killed, was later identified as Mohammed Azam. The police filed a case on behalf of Azam's family, police official Abdul Wakil Mengal said. It was not immediately clear what legal avenues the family can realistically pursue.