DocumentCode :
2132773
Title :
Convex optimization for exact rank recovery in topic models
Author :
Behmardi, Behrouz ; Raich, Raviv
Author_Institution :
Sch. of EECS, Oregon State Univ., Corvallis, OR, USA
fYear :
2011
fDate :
18-21 Sept. 2011
Firstpage :
1
Lastpage :
6
Abstract :
Topic models are widely used in a variety of applications including document classification and computer vision. The number of topics in the model plays an important role in terms of accuracy. We consider the problem of estimating the number of topics. In [1], a convex optimization approach was proposed to solve the problem via a constrained nuclear norm minimization. A standard semidefinite programming (SDP) was applied to solve the convex optimization only for a small size problem (e.g. 100× 100 matrix) due to its high computational complexity. To extend the applicability of the approach to large scale problems, we propose an accelerated gradient algorithm (AGA). Numerical results show that proposed algorithm can reliably solve a wide range of large scale problems in a shorter time than SDP solvers. Moreover, algorithms applied to a fairly large size real world dataset and results are provided.
Keywords :
computer vision; convex programming; document handling; gradient methods; image classification; statistical analysis; accelerated gradient algorithm; computer vision; constrained nuclear norm minimization; convex optimization; document classification; exact rank recovery; semidefinite programming; topic models; Accuracy; Computational complexity; Convergence; Convex functions; Minimization; Optimization; Sparse matrices; convex optimization; low rank matrix recovery; nuclear norm minimization; topic models;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Machine Learning for Signal Processing (MLSP), 2011 IEEE International Workshop on
Conference_Location :
Santander
ISSN :
1551-2541
Print_ISBN :
978-1-4577-1621-8
Electronic_ISBN :
1551-2541
Type :
conf
DOI :
10.1109/MLSP.2011.6064606
Filename :
6064606
Link To Document :
بازگشت