Diffusion LMS for multitask problems with overlapping hypothesis subspaces

Author

Jie Chen ; Richard, Cedric ; Hero, Alfred O. ; Sayed, Ali H.

Author_Institution

Univ. of Michigan, Ann Arbor, MI, USA

fYear

2014

fDate

21-24 Sept. 2014

Firstpage

1

Lastpage

6

Abstract

There are many important applications that are multitask-oriented in the sense that there are multiple optimum parameter vectors to be inferred simultaneously by networked agents. In this paper, we formulate an online multitask learning problem where node hypothesis spaces partly overlap. A cooperative algorithm based on diffusion adaptation is derived. Some results on its stability and convergence properties are also provided. Simulations are conducted to illustrate the theoretical results.

Keywords

convergence; learning (artificial intelligence); least mean squares methods; multi-agent systems; convergence properties; diffusion LMS; diffusion adaptation; multiple optimum parameter vectors; networked agents; node hypothesis spaces; online multitask learning problem; overlapping hypothesis subspaces; stability; Adaptive systems; Convergence; Estimation; Least squares approximations; Optimization; Signal processing algorithms; Vectors; Multitask learning; collaborative processing; diffusion strategy; distributed optimization;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning for Signal Processing (MLSP), 2014 IEEE International Workshop on

Conference_Location

Reims

Type

conf

DOI

10.1109/MLSP.2014.6958929

Filename

6958929