Learning spatially localized, parts-based representation

By S. Li, Xinwen Hou, HongJiang Zhang, Qiansheng Cheng. Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001, 2001

887 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/db53b8465a2d6b9dcba0a9c2ec7f10712c826306

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Fast Nonnegative Matrix Factorization: An Active-Set-Like Method and Comparisons (2011, background): “Due to the nonnegativity, the factors of low rank approximation give a natural interpretation: Each data item can be explained by an additive linear combination of physically-meaningful basis components [22].”
A flexible R package for nonnegative matrix factorization (2010, methodology): “For instance, the NMF package implements optimized C++ versions of the multiplicative updates from [4,10], which are used by several other NMF algorithms [13,24,31].”
Orthogonal nonnegative matrix t-factorizations for clustering (2006, background): “However, later experiments [16, 20] do not support the coherent part interpretation of NMF.”
Face recognition using Laplacianfaces (2005, methodology): “A common way to attempt to resolve this problem is to use dimensionality reduction techniques [1], [2], [8], [11], [12], [14], [22], [26], [28], [34], [37].”
Recognizing faces with PCA and ICA (2003, result): “Interestingly, this directly contradicts the claim made by Lee and Seung [27] (and echoed in [15, 20, 29]) that spatially localized feature vectors are better than overlapping (non spatially localized) feature vectors for face recognition, since spatially localized feature vectors performed much wor”
Nonnegative Matrix Factorization: An Analytical and Interpretive Tool in Computational Biology (2008, methodology): “Details of the algorithm are presented elsewhere [2,19,23,25,26,37,38,39].”
Constrained Nonnegative Matrix Factorization for Image Representation (2012, background): “In real applications such as image processing [2], face recognition [12], [25], and document clustering [13], [14], we usually set k ( m and k ( n.”
Sparse Representation for Computer Vision and Pattern Recognition (2010, methodology): “Component Analysis (PCA) approach of [67], Independent Component Analysis (ICA) [43], and Local Nonnegative Matrix Factorization (LNMF) [46].”
Nonnegative Matrix Factorization Based on Alternating Nonnegativity Constrained Least Squares and Active Set Method (2008, result): “Although the results presented in [20] show that the computed NMF generated parts-based basis vectors, the generation of a parts-based basis by the NMF depends on the data and the algorithm [14, 22].”
Sparse Nonnegative Matrix Factorization for Clustering (2008, result): “[15] showed examples that NMF generated holistic representations rather than part-based representations contrary to some earlier observations, and additional constraints such as sparsity on W or H [11] were recommended in order to obtain a sparser basis or representation.”
A Deep Matrix Factorization Method for Learning Attribute Representations (2015, methodology): “Although this method has been used by various previous works [37], [38] using the NMF model, it does not guarantee the non-negativity of h .”
On reduced rank nonnegative matrix factorization for symmetric nonnegative matrices (2004, background): “In [8,9], Lee and Seung suggested two useful algorithms which are easy to implement and have been successfully employed in many areas of applications, such as image processing, text information retrieval, and machine learning (see, for instance, [8,10,14]).”
Improving molecular cancer class discovery through sparse non-negative matrix factorization (2005, positive): “Thus Li et al. (2001) proposed local NMF (LNMF) with additional constraints to enforce the sparseness of the decomposition.”
Partial Face Recognition: Alignment-Free Approach (2013, background): “S. Liao and S.Z. Li are with the National Laboratory of Pattern Recognition and the Center for Biometrics and Security Research, Institute of Automation, Chinese Academy of Sciences, 1218 Automation Building, 95 Zhongguancun East Road, Beijing 100190, China.”