TY - GEN
T1 - Tensor decomposition via joint matrix schur decomposition
AU - Colombo, Nicolo
AU - Vlassis, Nikos
PY - 2016/6/19
Y1 - 2016/6/19
N2 - We describe an approach to tensor decomposition that involves extracting a set of observable matrices from the tensor and applying an approximate joint Schur decomposition on those matrices, and wc establish the corresponding first-order perturbation bounds. We develop a novel iterative Gauss-Newton algorithm for joint matrix Schur decomposition, which minimizes a nonconvex objective over the manifold of orthogonal matrices, and which is guaranteed to con-verge to a global optimum under certain conditions. We empirically demonstrate that our algorithm is faster and at least as accurate and robust than state-of-the-art algorithms for this problem.
AB - We describe an approach to tensor decomposition that involves extracting a set of observable matrices from the tensor and applying an approximate joint Schur decomposition on those matrices, and wc establish the corresponding first-order perturbation bounds. We develop a novel iterative Gauss-Newton algorithm for joint matrix Schur decomposition, which minimizes a nonconvex objective over the manifold of orthogonal matrices, and which is guaranteed to con-verge to a global optimum under certain conditions. We empirically demonstrate that our algorithm is faster and at least as accurate and robust than state-of-the-art algorithms for this problem.
UR - http://www.scopus.com/inward/record.url?scp=84998780357&partnerID=8YFLogxK
U2 - 10.5555/3045390.3045687
DO - 10.5555/3045390.3045687
M3 - Conference contribution
AN - SCOPUS:84998780357
T3 - 33rd International Conference on Machine Learning, ICML 2016
SP - 2820
EP - 2828
BT - ICML'16
A2 - Weinberger, Kilian Q.
A2 - Balcan, Maria Florina
PB - ACM
T2 - 33rd International Conference on Machine Learning, ICML 2016
Y2 - 19 June 2016 through 24 June 2016
ER -