Existing monitoring tools for multivariate data are often asymptotically distribution-free, computationally intensive, or require a large stretch of stable data. Many of these methods are not applicable to 'high dimension, low sample size' scenarios. With rapid technological advancement, high-dimensional data has become omnipresent in industrial applications. We propose a distribution-free change point monitoring method applicable to high dimensional data. Through an extensive simulation study, performance comparison has been done for different parameter values, under different multivariate distributions with complex dependence structures. The proposed method is robust and efficient in detecting change points under a wide range of shifts in the process distribution. A real-life application illustrated with the help of high-dimensional image surveillance dataset.

翻译：现有针对多变量数据的监控工具通常渐近无分布、计算密集或需要大量稳定数据段。许多此类方法不适用于"高维小样本"场景。随着技术的快速进步，高维数据在工业应用中已经无处不在。我们提出了一种适用于高维数据的无分布变点监测方法。通过广泛的模拟研究，针对不同参数值、具有复杂依赖结构的多变量分布进行了性能比较。该方法在过程分布的各类偏移下均具有鲁棒性和高效的变点检测能力。通过高维图像监控数据集的实际应用案例进行了验证。