TY - JOUR
T1 - Sparse non-negative tensor factorization using columnwise coordinate descent
AU - Liu, Ji
AU - Liu, Jun
AU - Wonka, Peter
AU - Ye, Jieping
N1 - Funding Information:
This work was supported by NSF IIS-0612069, IIS-0812551, CCF-0811790, NGA HM1582-08-1-0016, NSFC 60905035 and NSFC 61035003.
PY - 2012/1
Y1 - 2012/1
N2 - Many applications in computer vision, biomedical informatics, and graphics deal with data in the matrix or tensor form. Non-negative matrix and tensor factorization, which extract data-dependent non-negative basis functions, have been commonly applied for the analysis of such data for data compression, visualization, and detection of hidden information (factors). In this paper, we present a fast and flexible algorithm for sparse non-negative tensor factorization (SNTF) based on columnwise coordinate descent (CCD). Different from the traditional coordinate descent which updates one element at a time, CCD updates one column vector simultaneously. Our empirical results on higher-mode images, such as brain MRI images, gene expression images, and hyperspectral images show that the proposed algorithm is 12 orders of magnitude faster than several state-of-the-art algorithms.
AB - Many applications in computer vision, biomedical informatics, and graphics deal with data in the matrix or tensor form. Non-negative matrix and tensor factorization, which extract data-dependent non-negative basis functions, have been commonly applied for the analysis of such data for data compression, visualization, and detection of hidden information (factors). In this paper, we present a fast and flexible algorithm for sparse non-negative tensor factorization (SNTF) based on columnwise coordinate descent (CCD). Different from the traditional coordinate descent which updates one element at a time, CCD updates one column vector simultaneously. Our empirical results on higher-mode images, such as brain MRI images, gene expression images, and hyperspectral images show that the proposed algorithm is 12 orders of magnitude faster than several state-of-the-art algorithms.
KW - Columnwise coordinate descent
KW - Non-negative
KW - Sparse
KW - Tensor factorization
UR - http://www.scopus.com/inward/record.url?scp=80052781908&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80052781908&partnerID=8YFLogxK
U2 - 10.1016/j.patcog.2011.05.015
DO - 10.1016/j.patcog.2011.05.015
M3 - Article
AN - SCOPUS:80052781908
SN - 0031-3203
VL - 45
SP - 649
EP - 656
JO - Pattern Recognition
JF - Pattern Recognition
IS - 1
ER -