### Citations

6460 |
Neural networks and pattern recognition
- Bishop
- 1995
(Show Context)
Citation Context ...We compare the proposed algorithm with two representative feature selection algorithms: Laplacian Score [16] is a recent spectral graph-based unsupervised feature selection algorithm and Fisher Score =-=[4]-=- is a popular supervised feature selection algorithm which is employed in [16] for comparison. We implement sSelect algorithm in the Matlab environment. All experiments were conducted on a PENTIUM IV ... |

3721 | Normalized cuts and image segmentation
- Shi, Malik
(Show Context)
Citation Context ... learning results in spectral clustering algorithms [24, 10], which have been proved to be effective in many applications [3]. Spectral clustering algorithms, such as ratio cut [5] and normalized cut =-=[27]-=-, transform the original clustering problem to the cut problems on graph models. And the (local) optimal cluster indicator can be reconstructed from the eigenvectors of the corresponding matrix define... |

1673 | On spectral clustering: Analysis and an algorithm
- Ng, Jordan, et al.
- 2001
(Show Context)
Citation Context ...definite. 3. ∀x ∈ R n , x T Lx = � {vi,vj}∈E wij(xi − xj) 2 4. D · e = d and e T · D · e = volV .sApplying the spectral graph theory to unsupervised learning results in spectral clustering algorithms =-=[24, 10]-=-, which have been proved to be effective in many applications [3]. Spectral clustering algorithms, such as ratio cut [5] and normalized cut [27], transform the original clustering problem to the cut p... |

1515 | Wrappers for feature subset selection
- Kohavi, John
- 1997
(Show Context)
Citation Context ...X is whether labeled or not the feature selection algorithms can be either supervised or unsupervised [22, 12]. Supervised feature selection methods [8] can be broadly categorized into wrapper models =-=[19, 18]-=- and filter models [15, 26]. The wrapper model uses the predictive accuracy of a predetermined learning algorithm to determine the quality of selected features. These methods are prohibitively expensi... |

1490 |
Spectral graph theory
- Chung
- 1997
(Show Context)
Citation Context ...e task of learning from mixed labeled and unlabeled data is of semi-supervised learning [6]. In this paper, we present a semi-supervised feature selection algorithm based on the spectral graph theory =-=[7]-=-. The algorithm ranks features through a regularization framework, in which a feature’s relevance is evaluated by its fitness with both labeled and unlabeled data. The remainder of the paper is organi... |

1274 | An introduction to variable and feature selection
- Guyon, Elisseeff
- 2003
(Show Context)
Citation Context ...ne Learning, Spectral Analysis 1 Introduction The high dimensionality of data poses a challenge to learning tasks. In the presence of many irrelevant features, learning algorithms tend to overfitting =-=[14]-=-. Various studies show that features can be removed without performance deterioration [8]. Feature selection is one effective means to identify relevant features for dimension reduction [14, 22]. The ... |

1018 | Text classification from labelled and unlabelled documents using EM
- Nigam, McCallum, et al.
- 2000
(Show Context)
Citation Context ...beled data, or interpret it as unsupervised learning guided by constraints formed from labeled data. Generally the algorithms for semisupervised learning fall into three categories: Generative Models =-=[25, 2]-=-, Low Density Separation [21, 13] and Graph-Based Methods [31, 30]. The paucity of labeled data in semi-supervised learning requires that inductive bias [23] has to be introduced to make the learning ... |

725 | Semi-supervised learning using gaussian fields and harmonic functions
- Zhu, Ghahramani, et al.
(Show Context)
Citation Context ...nstraints formed from labeled data. Generally the algorithms for semisupervised learning fall into three categories: Generative Models [25, 2], Low Density Separation [21, 13] and Graph-Based Methods =-=[31, 30]-=-. The paucity of labeled data in semi-supervised learning requires that inductive bias [23] has to be introduced to make the learning possible. Inductive bias is some prior assumption(s). In semi-supe... |

652 | Learning with local and global consistency
- Zhou, Bousquet, et al.
(Show Context)
Citation Context ...group data. The three data sets are: (1) Pc vs. Mac (PCMAC), (2) Baseball vs. Hockey (HOCKEYBASE) and (3) Mac vs. Baseball (MACBASE). The four topics addressed in the three data sets are also used in =-=[30, 32]-=- and are widely used for performance evaluation for learning algorithms. The three data sets are generated from the version 20-news-18828. The articles in four topics were processed by the TMG softwar... |

455 | Feature selection: evaluation, application, and small sample performance
- Jain, Zongker
- 1997
(Show Context)
Citation Context ...to obtain. It is common to have a data set with huge dimensionality but small labeled-sample size. The data sets of this kind present a serious challenge, the so-called “small labeled-sample problem” =-=[1]-=-, to supervised feature selection, that is, when the labeled sample size is too small to carry sufficient information about the target concept, supervised feature selection algorithms fail with either... |

284 | Feature selection for classification
- Dash, Liu
- 1997
(Show Context)
Citation Context ...lenge to learning tasks. In the presence of many irrelevant features, learning algorithms tend to overfitting [14]. Various studies show that features can be removed without performance deterioration =-=[8]-=-. Feature selection is one effective means to identify relevant features for dimension reduction [14, 22]. The training data used in feature selection can be either labeled or unlabeled, corresponding... |

255 | Feature selection for discrete and numeric class machine learning
- Hall
- 1999
(Show Context)
Citation Context ... the feature selection algorithms can be either supervised or unsupervised [22, 12]. Supervised feature selection methods [8] can be broadly categorized into wrapper models [19, 18] and filter models =-=[15, 26]-=-. The wrapper model uses the predictive accuracy of a predetermined learning algorithm to determine the quality of selected features. These methods are prohibitively expensive to run for data with a l... |

252 | Toward integrating feature selection algorithms for classification and clustering
- Liu, Yu
(Show Context)
Citation Context ...erfitting [14]. Various studies show that features can be removed without performance deterioration [8]. Feature selection is one effective means to identify relevant features for dimension reduction =-=[14, 22]-=-. The training data used in feature selection can be either labeled or unlabeled, corresponding to supervised and unsupervised feature selection. In supervised feature selection [8, 14], feature relev... |

171 |
Spectral k-way ratio-cut partitioning and clustering
- Chan, Schlag, et al.
- 1993
(Show Context)
Citation Context ... theory to unsupervised learning results in spectral clustering algorithms [24, 10], which have been proved to be effective in many applications [3]. Spectral clustering algorithms, such as ratio cut =-=[5]-=- and normalized cut [27], transform the original clustering problem to the cut problems on graph models. And the (local) optimal cluster indicator can be reconstructed from the eigenvectors of the cor... |

153 | On the equivalence of nonnegative matrix factorization and spectral clustering
- Ding, He, et al.
- 2005
(Show Context)
Citation Context ...definite. 3. ∀x ∈ R n , x T Lx = � {vi,vj}∈E wij(xi − xj) 2 4. D · e = d and e T · D · e = volV .sApplying the spectral graph theory to unsupervised learning results in spectral clustering algorithms =-=[24, 10]-=-, which have been proved to be effective in many applications [3]. Spectral clustering algorithms, such as ratio cut [5] and normalized cut [27], transform the original clustering problem to the cut p... |

138 | Feature Selection for Unsupervised Learning
- Dy, Brodley
(Show Context)
Citation Context ...rvised and unsupervised feature selection. In supervised feature selection [8, 14], feature relevance can be evaluated by their correlation with the class label. And in unsupervised feature selection =-=[11, 12]-=-, without ∗ Technical Report, TR-06-022, Computer Science and Engineering (CSE) Department, Arizona State University (ASU), Tempe, AZ, 85281. {zheng.zhao, huan.liug}@asu.edu Zheng Zhao ∗ Huan Liu ∗ la... |

120 | Combining active learning and semisupervised learning using gaussian fields and harmonic functions
- Zhu, Ghahramani, et al.
- 2003
(Show Context)
Citation Context ...group data. The three data sets are: (1) Pc vs. Mac (PCMAC), (2) Baseball vs. Hockey (HOCKEYBASE) and (3) Mac vs. Baseball (MACBASE). The four topics addressed in the three data sets are also used in =-=[30, 32]-=- and are widely used for performance evaluation for learning algorithms. The three data sets are generated from the version 20-news-18828. The articles in four topics were processed by the TMG softwar... |

118 | Simultaneous feature selection and clustering using mixture models
- Law, Figueiredo, et al.
- 2004
(Show Context)
Citation Context ...e training data such as distance, consistency, dependency, information, and correlation [22]. Recently, an increasing number of researchers paid attention to developing unsupervised feature selection =-=[16, 20]-=-. It is a more loosely constrained search problem without class labels, depending on clustering quality measures [11], and can eventuate many equally valid feature subsets. The key issue of unsupervis... |

102 | Laplacian score for feature selection
- He, Cai, et al.
- 2006
(Show Context)
Citation Context ...rformance of sSelect. 4 Empirical Study We now empirically evaluate the performance of sSelect. We compare the proposed algorithm with two representative feature selection algorithms: Laplacian Score =-=[16]-=- is a recent spectral graph-based unsupervised feature selection algorithm and Fisher Score [4] is a popular supervised feature selection algorithm which is employed in [16] for comparison. We impleme... |

99 | Semi-supervised learning by entropy minimization
- Bengio, Grandvalet
- 2005
(Show Context)
Citation Context ...nsupervised learning guided by constraints formed from labeled data. Generally the algorithms for semisupervised learning fall into three categories: Generative Models [25, 2], Low Density Separation =-=[21, 13]-=- and Graph-Based Methods [31, 30]. The paucity of labeled data in semi-supervised learning requires that inductive bias [23] has to be introduced to make the learning possible. Inductive bias is some ... |

77 | Feature selection in unsupervised learning via evolutionary search
- Kim, Street, et al.
- 2000
(Show Context)
Citation Context ...X is whether labeled or not the feature selection algorithms can be either supervised or unsupervised [22, 12]. Supervised feature selection methods [8] can be broadly categorized into wrapper models =-=[19, 18]-=- and filter models [15, 26]. The wrapper model uses the predictive accuracy of a predetermined learning algorithm to determine the quality of selected features. These methods are prohibitively expensi... |

72 | A unified view of kernel k-means, spectral clustering and graph cuts
- Dhillon, Guan, et al.
- 2005
(Show Context)
Citation Context ...nal clustering problem to the cut problems on graph models. And the (local) optimal cluster indicator can be reconstructed from the eigenvectors of the corresponding matrix defined in the cut problem =-=[9]-=-. Instead of reconstructing the cluster indicators from eigenvectors, we show a way to construct them from feature vectors. By doing so, the fitness of cluster indicators can be evaluated by both labe... |

57 | Unsupervised Feature Selection Applied to ContentBased Retrieval of Lung Images
- Dy, Brodley, et al.
- 2003
(Show Context)
Citation Context ...rvised and unsupervised feature selection. In supervised feature selection [8, 14], feature relevance can be evaluated by their correlation with the class label. And in unsupervised feature selection =-=[11, 12]-=-, without ∗ Technical Report, TR-06-022, Computer Science and Engineering (CSE) Department, Arizona State University (ASU), Tempe, AZ, 85281. {zheng.zhao, huan.liug}@asu.edu Zheng Zhao ∗ Huan Liu ∗ la... |

54 |
Semi-supervised learning via gaussian processes
- Lawrence, Jordan
- 2004
(Show Context)
Citation Context ...nsupervised learning guided by constraints formed from labeled data. Generally the algorithms for semisupervised learning fall into three categories: Generative Models [25, 2], Low Density Separation =-=[21, 13]-=- and Graph-Based Methods [31, 30]. The paucity of labeled data in semi-supervised learning requires that inductive bias [23] has to be introduced to make the learning possible. Inductive bias is some ... |

40 | Tmp: a matlab toolbox for generating term-document matrices from text collections
- Zeimpekis, Gallopoulos
- 2006
(Show Context)
Citation Context ...idely used for performance evaluation for learning algorithms. The three data sets are generated from the version 20-news-18828. The articles in four topics were processed by the TMG software package =-=[29]-=- with the following options: (1) passing all words through the Porter stemmer before counting them; (2) tossing out any token which is on the stoplist; and (3) ignoring words that occur in 4 or fewer ... |

27 |
Semi-supervised Clustering: Probabilistic Models, Algorithms and Experiments
- Basu
- 2005
(Show Context)
Citation Context ...beled data, or interpret it as unsupervised learning guided by constraints formed from labeled data. Generally the algorithms for semisupervised learning fall into three categories: Generative Models =-=[25, 2]-=-, Low Density Separation [21, 13] and Graph-Based Methods [31, 30]. The paucity of labeled data in semi-supervised learning requires that inductive bias [23] has to be introduced to make the learning ... |

25 |
Theoretical and empirical analysis of Relief and ReliefF
- Robnik-Sikonja, Kononenko
(Show Context)
Citation Context ... the feature selection algorithms can be either supervised or unsupervised [22, 12]. Supervised feature selection methods [8] can be broadly categorized into wrapper models [19, 18] and filter models =-=[15, 26]-=-. The wrapper model uses the predictive accuracy of a predetermined learning algorithm to determine the quality of selected features. These methods are prohibitively expensive to run for data with a l... |

24 | IDR/QR: An incremental dimension reduction algorithm via QR decomposition - Ye, Li, et al. - 2004 |

3 | A combinatorial view of the graph Laplacians
- Huang
- 2005
(Show Context)
Citation Context ...tering Indicator Construction The normalized min-cut clustering algorithm was first proposed by Shi and Malik in [27], and has been shown to be superior to other cluster algorithms, such as ratio cut =-=[17]-=-. Our method resorts to transforming feature vectors to the cluster indicators of normalized min-cut. Given a graph G = (V, E) constructed from data X, the normalized min-cut clustering algorithm find... |

2 | On spectral properties of graphs and their application to clustering
- Bilu
(Show Context)
Citation Context ... d and e T · D · e = volV .sApplying the spectral graph theory to unsupervised learning results in spectral clustering algorithms [24, 10], which have been proved to be effective in many applications =-=[3]-=-. Spectral clustering algorithms, such as ratio cut [5] and normalized cut [27], transform the original clustering problem to the cut problems on graph models. And the (local) optimal cluster indicato... |