Title :
A Semi-Definite Programming Embedding Framework for Local Preserving Manifold Learning
Author :
Zeng, Xianhua ; Gan, Ling ; Wang, Jian
Author_Institution :
Coll. of Comput. Sci. & Technol., Chongqing Univ. of Posts & Telecommun., Chongqing, China
Abstract :
A semi-definite programming embedding framework is presented for local preserving manifold learning in this paper. Under the framework, three unstable algorithms (LE, LLE and LTSA) are respectively converted into the stable semi-definite programming embedding algorithms (named as SDPE-LE, SDPE-LLE and SDPE-LTSA). The advantages and effectiveness of these new algorithms are demonstrated via the experimental results on synthetic dataset and real image dataset.
Keywords :
learning (artificial intelligence); nonlinear programming; visual databases; local preserving manifold learning; real image dataset; semidefinite programming embedding framework; synthetic dataset; unstable algorithms; Laplace equations; Manifolds; Optimization methods; Programming; Software; Telecommunications;
Conference_Titel :
Pattern Recognition (CCPR), 2010 Chinese Conference on
Conference_Location :
Chongqing
Print_ISBN :
978-1-4244-7209-3
Electronic_ISBN :
978-1-4244-7210-9
DOI :
10.1109/CCPR.2010.5659162
