Title: A robust principal component analysis for outlier identification in messy microcalorimeter data Show Chinese title

Title: 一种用于杂乱微量热计数据中异常值识别的鲁棒主成分分析

Abstract: A principal component analysis (PCA) of clean microcalorimeter pulse records can be a first step beyond statistically optimal linear filtering of pulses towards a fully non-linear analysis. For PCA to be practical on spectrometers with hundreds of sensors, an automated identification of clean pulses is required. Robust forms of PCA are the subject of active research in machine learning. We examine a version known as coherence pursuit that is simple, fast, and well matched to the automatic identification of outlier records, as needed for microcalorimeter pulse analysis. Show Chinese abstract

Abstract: 主成分分析（PCA）对干净的微热量计脉冲记录进行分析，可以作为超越统计最优线性滤波的初步步骤，进而实现完全非线性的分析。 为了在拥有数百个传感器的光谱仪上实际应用PCA，需要自动识别干净的脉冲。 鲁棒的PCA形式是机器学习领域活跃的研究课题。 我们研究了一种称为相干性追求的方法，这种方法简单、快速，并且非常适合自动识别异常记录，如微热量计脉冲分析所需的那样。