Fast and stable recovery of Approximately low multilinear rank tensors from multi-way compressive measurements

We introduce a reconstruction formula that allows one to recover an N-order tensor X ϵ RI1×...×In from a reduced set of multi-way compressive measurements by exploiting its low multilinear rank structure. It is proved that, in the matrix case (N = 2), the proposed reconstruction is stable in the sense that the approximation error is proportional to the one provided by the best low-rank approximation, i.e ||X - X||2 ≤ K||X - X0||2, where K is a constant and X0 is the corresponding truncated SVD of X. We also present simulation results indicating that the same stable behavior is observed with higher order tensors (N > 2). In addition, it is shown that, an interesting property of multi-way measurements allows us to build the reconstruction based on compressive linear measurements of fibers taken only in two selected modes, independently of the tensor order N. Simulation results using real-world 2D and 3D signals are presented illustrating our results and comparing the reconstructions against the best low multilinear rank approximations and the reconstructions obtained by using the Kronecker-CS approach.

[1]  Andrzej Cichocki,et al.  Block sparse representations of tensors using Kronecker bases , 2012, 2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[2]  Jieping Ye,et al.  Tensor Completion for Estimating Missing Values in Visual Data , 2013, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Kinjiro Amano,et al.  Information limits on neural identification of colored surfaces in natural scenes , 2004, Visual Neuroscience.

[4]  Paul Hasler,et al.  Compressive Sensing on a CMOS Separable-Transform Image Sensor , 2010, Proc. IEEE.

[5]  Andrzej Cichocki,et al.  Computing Sparse Representations of Multidimensional Signals Using Kronecker Bases , 2013, Neural Computation.

[6]  H. Howard Fan,et al.  Low rank variational tensor recovery for multi-linear inverse problems , 2011, 2011 Conference Record of the Forty Fifth Asilomar Conference on Signals, Systems and Computers (ASILOMAR).

[7]  Tamara G. Kolda,et al.  Tensor Decompositions and Applications , 2009, SIAM Rev..

[8]  Daniel Kressner,et al.  A literature survey of low‐rank tensor approximation techniques , 2013, 1302.7121.

[9]  Dave Zachariah,et al.  Alternating Least-Squares for Low-Rank Matrix Reconstruction , 2012, IEEE Signal Processing Letters.

[10]  Reinhold Schneider,et al.  Low rank tensor recovery via iterative hard thresholding , 2016, ArXiv.

[11]  Adrian Stern,et al.  Compressed Imaging With a Separable Sensing Operator , 2009, IEEE Signal Processing Letters.

[12]  Joos Vandewalle,et al.  A Multilinear Singular Value Decomposition , 2000, SIAM J. Matrix Anal. Appl..

[13]  Richard G. Baraniuk,et al.  Kronecker Compressive Sensing , 2012, IEEE Transactions on Image Processing.

[14]  Emmanuel J. Candès,et al.  Matrix Completion With Noise , 2009, Proceedings of the IEEE.

[15]  Babak Hossein Khalaj,et al.  A unified approach to sparse signal processing , 2009, EURASIP Journal on Advances in Signal Processing.

[16]  Dan Schonfeld,et al.  Generalized tensor compressive sensing , 2013, 2013 IEEE International Conference on Multimedia and Expo (ICME).

[17]  Anastasios Kyrillidis,et al.  Multi-Way Compressed Sensing for Sparse Low-Rank Tensors , 2012, IEEE Signal Processing Letters.

[18]  Tamara G. Kolda,et al.  Scalable Tensor Factorizations for Incomplete Data , 2010, ArXiv.

[19]  Andrzej Cichocki,et al.  Multidimensional compressed sensing and their applications , 2013, WIREs Data Mining Knowl. Discov..