We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
An Automatic Model Selection‐Based Machine Learning Framework to Estimate FORC Distributions.
- Authors
Heslop, D.; Roberts, A. P.; Oda, H.; Zhao, X.; Harrison, R. J.; Muxworthy, A. R.; Hu, P.‐X.; Sato, T.
- Abstract
First‐order reversal curve (FORC) distributions are a powerful diagnostic tool for characterizing and quantifying magnetization processes in fine magnetic particle systems. Estimation of FORC distributions requires the computation of the second‐order mixed derivative of noisy magnetic hysteresis data. This operation amplifies measurement noise, and for weakly magnetic systems, it can compromise estimation of a FORC distribution. Previous processing schemes, which are based typically on local polynomial regression, have been developed to smooth FORC data to suppress detrimental noise. Importantly, the smoothed FORC distribution needs to be consistent with the measurement data from which it was estimated. This can be a challenging task even for expert users, who must adjust subjectively parameters that define the form and extent of smoothing until a "satisfactory" FORC distribution is obtained. For nonexpert users, estimation of FORC distributions using inappropriate smoothing parameters can produce distorted results corrupted by processing artifacts, which can lead to spurious inferences concerning the magnetic system under investigation. We have developed a statistical machine learning framework based on a probabilistic model comparison to guide the estimation of FORC distributions. An intuitive approach is presented that reveals regions of a FORC distribution that may have been smoothed inappropriately. An associated metric can also be used to compare data preparation and local regression schemes to assess their suitability for processing a given FORC data set. Ultimately, our approach selects FORC smoothing parameters in a probabilistic fashion, which automates the derivative estimation process regardless of user expertise. Key Points: We develop a probabilistic framework for estimating FORC distributionsUsing Bayesian model selection, we identify underfitting and overfitting of FORC distributionsFORCsensei software can be used to determine smoothing parameters to estimate a FORC distribution
- Subjects
MAGNETIC particles; MAGNETIC hysteresis; NOISE measurement; MACHINE learning; MAGNETIC properties of rocks
- Publication
Journal of Geophysical Research. Solid Earth, 2020, Vol 125, Issue 10, p1
- ISSN
2169-9313
- Publication type
Article
- DOI
10.1029/2020JB020418