NEWMA: a new method for scalable model-free online change-point detection

Abstract

We consider the problem of detecting abrupt changes in the distribution of a multidimensional time series, with limited computing power and memory. In this paper, we propose a new method for model-free online change-point detection that relies only on fast and light recursive statistics, inspired by the classical Exponential Weighted Moving Average algorithm (EWMA). The proposed idea is to compute two EWMA statistics on the stream of data with different forgetting factors, and to compare them. By doing so, we show that we implicitly compare recent samples with older ones, without the need to explicitly store them. Additionally, we leverage Random Features to efficiently use the Maximum Mean Discrepancy as a distance between distributions. We show that our method is orders of magnitude faster than usual non-parametric methods for a given accuracy.

Machine learning and the physical sciences

“…8 2a). Machine learning applications range from transfer learning for deep neural networks, time series analysis -with a feedback loop implementing so-called echo-state networks (Dong et al, 2018), or change-point detection (Keriven et al, 2018). For large-dimensional data, these devices already outperform CPUs or GPUs both in speed and power consumption.…”

Kernel Computations from Large-Scale Random Features Obtained by Optical Processing Units

“…Their results demonstrate that the OPU makes a significant contribution towards making kernel methods more practical for large-scale applications with the potential to drastically decrease computation time and memory, as well as power consumption. The OPU has also been applied to other frameworks like reservoir computing [12,13] and anomaly detection [14].…”