Multi-component background learning automates signal detection for spectroscopic data

作     者：Sebastian E.Ament Helge S.Stein Dan Guevarra Lan Zhou Joel A.Haber David A.Boyd Mitsutaro Umehara John M.Gregoire Carla P.Gomes 

作者机构：Department of Computer ScienceCornell UniversityIthacaNY 14850USA Joint Center for Artificial PhotosynthesisCalifornia Institute of TechnologyPasadenaCA 91125USA Future Mobility Research DepartmentToyota Research Institute of North AmericaAnn ArborMI 48105USA 

出 版 物：《npj Computational Materials》 (计算材料学（英文）)

年 卷 期：2019年第5卷第1期

页      面：484-490页

核心收录：

学科分类：0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 

基　　金：The development of the MCBL algorithm,inkjet printing synthesis,and Raman measurements were supported by a an Accelerated Materials Design and Discovery grant from the Toyota Research Institute Initial design of the algorithm and data procurement were supported by the NSF Expedition award for Computational Sustainability CCF-1522054 and by Army Research Office(ARO)award W911-NF-14-1-0498 The implementation of the algorithm for automated,unsupervised operation was supported by MURI/AFOSR grant FA9550 Compute infrastructure was provided by NSF award CNS-0832782 and by ARO DURIP award W911NF-17-1-0187 The sputter deposition and XRD measurements were supported through the Office of Science of the U.S.Department of Energy under Award No.DE-SC0004993 

主　　题：signal component spectroscopic 

摘      要：Automated experimentation has yielded data acquisition rates that supersede human processing *** Intelligence offers new possibilities for automating data interpretation to generate large,high-quality *** subtraction is a long-standing challenge,particularly in settings where multiple sources of the background signal coexist,and automatic extraction of signals of interest from measured signals accelerates data ***,we present an unsupervised probabilistic learning approach that analyzes large data collections to identify multiple background sources and establish the probability that any given data point contains a signal of *** approach is demonstrated on X-ray diffraction and Raman spectroscopy data and is suitable to any type of data where the signal of interest is a positive addition to the background *** the model can incorporate prior knowledge,it does not require knowledge of the signals since the shapes of the background signals,the noise levels,and the signal of interest are simultaneously learned via a probabilistic matrix factorization *** identification of interpretable signals by unsupervised probabilistic learning avoids the injection of human bias and expedites signal extraction in large datasets,a transformative capability with many applications in the physical sciences and beyond.