Online Distribution Shift Detection via Recency Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Luo, Rachel, Sinha, Rohan, Sun, Yixiao, Hindy, Ali, Zhao, Shengjia, Savarese, Silvio, Schmerling, Edward, Pavone, Marco
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917669195743232
author Luo, Rachel
Sinha, Rohan
Sun, Yixiao
Hindy, Ali
Zhao, Shengjia
Savarese, Silvio
Schmerling, Edward
Pavone, Marco
author_facet Luo, Rachel
Sinha, Rohan
Sun, Yixiao
Hindy, Ali
Zhao, Shengjia
Savarese, Silvio
Schmerling, Edward
Pavone, Marco
contents When deploying modern machine learning-enabled robotic systems in high-stakes applications, detecting distribution shift is critical. However, most existing methods for detecting distribution shift are not well-suited to robotics settings, where data often arrives in a streaming fashion and may be very high-dimensional. In this work, we present an online method for detecting distribution shift with guarantees on the false positive rate - i.e., when there is no distribution shift, our system is very unlikely (with probability $< ε$) to falsely issue an alert; any alerts that are issued should therefore be heeded. Our method is specifically designed for efficient detection even with high dimensional data, and it empirically achieves up to 11x faster detection on realistic robotics settings compared to prior work while maintaining a low false negative rate in practice (whenever there is a distribution shift in our experiments, our method indeed emits an alert). We demonstrate our approach in both simulation and hardware for a visual servoing task, and show that our method indeed issues an alert before a failure occurs.
format Preprint
id arxiv_https___arxiv_org_abs_2211_09916
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Online Distribution Shift Detection via Recency Prediction
Luo, Rachel
Sinha, Rohan
Sun, Yixiao
Hindy, Ali
Zhao, Shengjia
Savarese, Silvio
Schmerling, Edward
Pavone, Marco
Robotics
Machine Learning
When deploying modern machine learning-enabled robotic systems in high-stakes applications, detecting distribution shift is critical. However, most existing methods for detecting distribution shift are not well-suited to robotics settings, where data often arrives in a streaming fashion and may be very high-dimensional. In this work, we present an online method for detecting distribution shift with guarantees on the false positive rate - i.e., when there is no distribution shift, our system is very unlikely (with probability $< ε$) to falsely issue an alert; any alerts that are issued should therefore be heeded. Our method is specifically designed for efficient detection even with high dimensional data, and it empirically achieves up to 11x faster detection on realistic robotics settings compared to prior work while maintaining a low false negative rate in practice (whenever there is a distribution shift in our experiments, our method indeed emits an alert). We demonstrate our approach in both simulation and hardware for a visual servoing task, and show that our method indeed issues an alert before a failure occurs.
title Online Distribution Shift Detection via Recency Prediction
topic Robotics
Machine Learning
url https://arxiv.org/abs/2211.09916