A Robust Principal Component Analysis for Outlier Identification in Messy Microcalorimeter Data.

Area of Science:

• Physics
• Data Science
• Spectroscopy

Background:

• Microcalorimeter pulse analysis traditionally uses linear filtering.
• Advanced analysis requires moving beyond linear methods to explore nonlinear dynamics.
• Spectrometers with numerous sensors necessitate efficient data processing techniques.

Purpose of the Study:

• To explore principal component analysis (PCA) as a method for microcalorimeter pulse analysis.
• To investigate automated identification of clean pulse records for practical PCA implementation.
• To evaluate coherence pursuit as a robust PCA variant for outlier detection.

Main Methods:

• Application of principal component analysis (PCA) to microcalorimeter pulse records.
• Development and testing of automated methods for identifying clean pulses.
• Examination of coherence pursuit for its speed and suitability in outlier identification.

Main Results:

• PCA can serve as a precursor to nonlinear analysis of microcalorimeter pulses.
• Automated clean pulse identification is crucial for scalable PCA in spectrometers.
• Coherence pursuit demonstrates potential for efficient outlier record identification.

Conclusions:

• Principal component analysis (PCA) is a viable step towards nonlinear analysis of microcalorimeter data.
• Coherence pursuit offers a practical and efficient approach for automated outlier detection in pulse analysis.