Abstract
A new tensor approximation method is developed based on the CANDECOMP/PARAFAC (CP) factorization that enjoys both sparsity (i.e., yielding factor matrices with some non-zero elements) and resistance to outliers and non-Gaussian measurement noise. This method utilizes a robust bounded loss function for errors in the low-rank tensor approximation while encouraging sparsity with Lasso (or ℓ1-) regularization to the factor matrices (of a tensor data). A simple alternating, iteratively reweighted (IRW) Lasso algorithm is proposed to solve the resulting optimization problem. Simulation studies illustrate that the proposed method provides excellent performance in terms of mean square error accuracy for heavy-tailed noise conditions, with relatively small loss in conventional Gaussian noise.
Original language | English (US) |
---|---|
Title of host publication | 2014 IEEE Workshop on Statistical Signal Processing, SSP 2014 |
Publisher | IEEE Computer Society |
Pages | 420-423 |
Number of pages | 4 |
ISBN (Print) | 9781479949755 |
DOIs | |
State | Published - 2014 |
Event | 2014 IEEE Workshop on Statistical Signal Processing, SSP 2014 - Gold Coast, QLD, Australia Duration: Jun 29 2014 → Jul 2 2014 |
Publication series
Name | IEEE Workshop on Statistical Signal Processing Proceedings |
---|
Other
Other | 2014 IEEE Workshop on Statistical Signal Processing, SSP 2014 |
---|---|
Country | Australia |
City | Gold Coast, QLD |
Period | 6/29/14 → 7/2/14 |
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering
- Applied Mathematics
- Signal Processing
- Computer Science Applications
Keywords
- Iteratively reweighted least squares
- Lasso
- big data
- regularization
- robust loss function