Dissertation > Excellent graduate degree dissertation topics show
Nonnegative Matrix Factorization with Sparseness Constraint
Author: ZhangYuFei
Tutor: ChengMingSong
School: Dalian University of Technology
Course: Computational Mathematics
Keywords: Non - negative matrix factorization The steepest descent method Auxiliary function Sparsity Constraint KKT conditions Image storage
CLC: O151.21
Type: Master's thesis
Year: 2010
Downloads: 207
Quote: 1
Read: Download Dissertation
Abstract
|
Non- negative matrix factorization (Nonnegative Matrix Factorization, NMF) is a matrix decomposition method recently proposed international . Compared with other matrix factorization , NMF special about its by the introduction of non- negative constraint matrix decomposition process , this constraint original data will be based on the part of the representation , in order to better reflect the local features of the original data . Due to a large number of experimental data in the real world are non- negative expression , which makes the presence of a wide range of applications , such as blind source separation and non negative signal analysis , pattern recognition , text knowledge mining , digital image watermark as well as facial expression recognition [16 - 21] and other fields . Depending on the application , the decomposition of the matrix elements can have different interpretations . For example , the method described by Lee and Seung [ '] is to face image decomposed into portions ( such as the lips , eyes , nose , ears, etc. ). The first part of this paper describes the background of non- negative matrix factorization , the description of the problem , and Research , NMF problem there described the shortcomings and difficulties . The second part first introduced this article related to prior knowledge , the next listed some of the NMF algorithm , and its details , making you a more in-depth understanding of the NMF method . The third part of the core of this article : plus the sparse constraint of non- negative matrix decomposition algorithm , the algorithm is an improvement of the previous NMF algorithm . NMF 's purpose is to the original non- negative matrix V ∈ Rm × n decomposed into the product of the new non - negative matrix W ∈ Rm × r and H ∈ Rr x n : V ≈ WH , wherein r satisfies : ( Mn ) r lt ; Mn and the decomposition error as small as possible . Our goal is to ensure on the basis of the non - negative constraint conditions and the decomposition of accuracy by sparse constraint conditions on the increase in the objective function , obtained in the decomposition matrix as sparse as possible , to save storage space more . The fourth part is the numerical experiments , the use of numerical examples to verify the effectiveness of the correctness of the conclusions in the paper as well as solving methods , and by different methods to verify the algorithm does have a fast convergence rate and good sparsity .
|
Related Dissertations
- Research on Co-clustering and Application,TP311.13
- Non-negative Local Coordinate Factorization for Image Representation,TP391.41
- Non-negative matrix factorization based on sparse image retrieval,TP391.41
- Hyperspectral Unmixing Based on Nonnegative Matrix Factorization,TP751.1
- A Non-negative Matrix Factorization Model with Sparse Constraint and Its Initialization,O151.21
- The new incremental learning algorithm for support vector machines,TP181
- Technology Combining Fingerprint Recognition with Watermarking Used for Copyright Protection and Authentication,TP309.7
- Face Recognation Method Based on Non-negative Matrix Factorization with BP Neural Networks,TP391.41
- P-Continuous Method for P-Laplacian,O241.82
- Write photorefractive crystal optical modulation wave coupling and image storage in two,O734
- Research and Design of PACS Workstation and Relative Technology Based on DICOM3.0,R-39
- The Improvement of Genetic Algorithm and Its Application on Optimal Method of Urban Traffic Signal Control,TP18
- Facial Expression Recognition based on non - negative matrix decomposition,TP391.41
- Research on Face Recognition Methods with Subspace Linear Projections,TP391.41
- Digital Image Watermarking Algorithm Based on Wavelet Transform and the NMF,TP309.7
- The Algorithm Research on Nonlinear Optimization Problems,O221.2
- Study on Model-Based Fragmentation in OPC,TN305.7
- Research on Face Feature Subspace Methods,TP391.41
- Global Optimization of the auxiliary function method,O224
- Constraint -based sparse NMF algorithm for face recognition Aging Simulation and its application to face recognition,TP391.41
CLC: > Mathematical sciences and chemical > Mathematics > Algebra,number theory, portfolio theory > Theory of algebraic equations,linear algebra > Linear Algebra > Matrix theory
© 2012 www.DissertationTopic.Net Mobile
|