ATT: Adaptive Algorithms for Tensor Train Decomposition of Streaming Tensors

We propose a novel adaptive algorithm for TT decomposition of streaming tensors whose slices are serially acquired over time. By leveraging the alternating minimization framework, our estimator minimizes an exponentially weighted least-squares cost function in an efficient way.

Dependencies

• Our code requires the Tensor Toolbox which is already attached in this repository.
• MATLAB R2019a

Demo

Quick Start: Just run the file DEMO.m

Some Results

Effect of the noise level and time-varying factors on the performance of our method

Performance of three TT decomposition algorithms in a time-varying scenario

Reference:

This code is free and open source for research purposes. If you use this code, please acknowledge the following paper.

[1] L.T. Thanh, K. Abed-Meraim, N.L. Trung, and R Boyer. "Adaptive Algorithms for Tensor Train Decomposition of Streaming Tensors". European Signal Processing Conference (EUSIPCO), 2020.