By Topic

Machine Learning and Cybernetics (ICMLC), 2011 International Conference on

Date 10-13 July 2011

Go

Filter Results

Displaying Results 1 - 25 of 98
  • [Front and back cover]

    Publication Year: 2011 , Page(s): c1 - c4
    Save to Project icon | Request Permissions | PDF file iconPDF (872 KB)  
    Freely Available from IEEE
  • [Front matter]

    Publication Year: 2011 , Page(s): 1 - 2
    Save to Project icon | Request Permissions | PDF file iconPDF (84 KB)  
    Freely Available from IEEE
  • Greetings from the General Chairs

    Publication Year: 2011 , Page(s): i - iii
    Save to Project icon | Request Permissions | PDF file iconPDF (49 KB)  
    Freely Available from IEEE
  • Organizing Committee

    Publication Year: 2011 , Page(s): iv - xi
    Save to Project icon | Request Permissions | PDF file iconPDF (66 KB)  
    Freely Available from IEEE
  • Contents

    Publication Year: 2011 , Page(s): 1 - 5
    Save to Project icon | Request Permissions | PDF file iconPDF (70 KB)  
    Freely Available from IEEE
  • Medical image retrieval: Multiple regression models for user's search target

    Publication Year: 2011 , Page(s): 1436 - 1443
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (174 KB) |  | HTML iconHTML  

    Breast cancer has been one of leading cancers in women around the world. A great number of digital mammograms are generated in hospitals and screening centers. Those digital mammograms can further be used for study and research by medical professionals. Content-based image retrieval refers to the retrieval of images whose contents are similar to a query example, using information derived from the images themselves. Relevance feedback, expressing the user's search target, can be used to bridge the semantic gap and improve the performance of CBIR systems. This study proposes a learning method for relevance feedback learning, which develops multiple logistic regression models to generalize the classification problem and provide an estimate of probability of class membership. To build the model, relevance feedback is utilized as the training data and the IRLS method is applied to estimate the parameters of the regression model and compute the maximum likelihood. Logistic regression models are created individually. After logistic regression models are fitted, discriminating features are selected by the measure of goodness of fit statistics. The weights of those discriminating features can be assigned according to their individual contributions to the maximum likelihood. The probability of the membership of the relevant class can therefore be obtained for each image of the database. Experimental results show that the proposed learning method can effectively improve the average precision from 30% to 65% through five iterations of relevance feedback rounds. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A new method of distance measure for graph-based semi-supervised learning

    Publication Year: 2011 , Page(s): 1444 - 1448
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1026 KB) |  | HTML iconHTML  

    With an intensive study of the existing density-sensitive distance measures, we proposed a new distance measure for graph-based semi-supervised learning. The proposed measure can not only effectively amplify the distance between data points in different high-density regions, but also reduce the distance among data points in a same high-density region. Then, a graph-based semi-supervised clustering algorithm is presented based on the proposed distance measure. Experimental results on some UCI data sets show that the proposed method has obvious advantages than the old one. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A new non-negative matrix factorization algorithm with sparseness constraints

    Publication Year: 2011 , Page(s): 1449 - 1452
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (241 KB) |  | HTML iconHTML  

    The non-negative matrix factorization (NMF) aims to find two matrix factors for a matrix X such that X ≈ W H, where W and H are both nonnegative matrices. The non-negativity constraint arises often naturally in applications in physics and engineering. In this paper, we propose a new NMF approach, which incorporates sparseness constraints explicitly. The new model can learn much sparser matrix factorization. Also, an objective function is defined to impose the sparseness constraint, in addition to the non-negative constraint. Experimental results on two document datasets show the effectiveness and efficiency of the proposed method. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A retrieval system of vehicles based on recognition of license plates

    Publication Year: 2011 , Page(s): 1453 - 1459
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (374 KB) |  | HTML iconHTML  

    It is a laborious process for people to search vehicles from a large amount of video volumes. We build a retrieval system of vehicles using a novel method which recognizes license plate from surveillance videos. The core idea is to automatically convert video information into text information such that the tracing problem of vehicles can be simplified since all records of vehicles are efficiently managed by a well-designed vehicle database. Using a vehicle license plate as a key word, the retrieval system not only finds completed information of the vehicle but also visualizes the track of the vehicle by an electrical map. Firstly we propose a new algorithm to detect the key frames including vehicle license plate characters from video. The most important part is to distinguish vehicles from other objects. Secondly, we correct vehicle plates using modified Radon Transform. Thirdly, some necessary information about vehicles like time stamps is also recorded in the vehicles database. The experimental results show that the proposed algorithm is effective to deal with videos and record vehicles information. Text and visualization information which the database provides can meet users' requirements for searching vehicles. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A similarity measure for text processing

    Publication Year: 2011 , Page(s): 1460 - 1465
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (698 KB) |  | HTML iconHTML  

    In this paper, we propose a novel similarity measure for document data processing. For two document vectors, the proposed measure takes three cases into account: a) The feature considered appears in both documents, b) the feature considered appears in only one document, and c) the feature considered appears in none of the documents. For the first case, we give a lower bound and decrease the similarity according to the difference between the feature values of the two documents. For the second case, we give a fixed value disregarding the magnitude of the feature value. For the last case, we treat it as an identity, Experimental results show that our proposed method can work more effectively than others. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Clustering web documents based on Multiclass spectral clustering

    Publication Year: 2011 , Page(s): 1466 - 1471
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (790 KB) |  | HTML iconHTML  

    Multiclass spectral clustering is a clustering method which has been successfully applied in image segmentation and many other aspects. In this paper, Multiclass spectral clustering is used to cluster web documents including both English and Chinese pages. Through experiments, we found that Multiclass spectral clustering can be well used in web document clustering, and the method not only works well to cluster English web documents but also works well to cluster Chinese web documents clustering. We applied our method to a web search engine, and users can get the suitable results easily by just selecting the desirable classes. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CSG-Tag: Constraint based Synchronous Grammar Tree Annotation System

    Publication Year: 2011 , Page(s): 1472 - 1478
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (236 KB) |  | HTML iconHTML  

    The construction of grammars and the acquisition of syntactic structures from corpora are always considered as a time consuming task. Moreover, according to the purpose of the application, different standards have to be defined. In Machine Translation (MT), the situation is even more complicated since it covers two languages. In this paper, CSG-Tag, a Constraint based Synchronous Grammar (CSG) Tree Annotation System is proposed. This system provides a semi-automatic annotation process in the creation of syntactic structure of the source sentence linked with the corresponding target sentential patterns. All learned information are stored in Extensible Markup Language (XML) format and can be converted into grammar rules in application to MT. Moreover, the system has a function to import monolingual skeletal bracketing syntactic tree and Translation Corresponding Tree (TCT) structures in the creation of CSG rules. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient top-k support documents for expert search using relationship in a social network

    Publication Year: 2011 , Page(s): 1479 - 1484
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (214 KB) |  | HTML iconHTML  

    Searching experts for helping make decision in an organization is an effective solution. Traditional approaches of expert search only use the expertise information of a single expert and ignore relationship between persons. Recently some research shows that relationship between persons is also helpful. However, most approaches cost much time and energy to establish expertise profiles for all experts and extract varies of social relationship between them. In this paper, we propose an approach which can not only efficiently collect expertise information of each expert through top-k support documents, but also use effective co-occurrence relationship between expert candidates to rank the target experts. The use of co-occurrence relationship aims to quickly build a social network. And it also enhances reliability of relevance between expert candidates and a given topic, and improves accuracy of recommended experts. Experimental results on W3C collection show that our approach outperforms the baseline approaches. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Extracting Chinese abbreviation-definition pairs from anchor texts

    Publication Year: 2011 , Page(s): 1485 - 1491
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (192 KB) |  | HTML iconHTML  

    This paper proposes an automatic scheme to extract Chinese abbreviations and their corresponding definitions from large-scale anchor texts. This method is motivated by the observation that the more frequently two anchor texts point to the same web page, the more related they are. Since abbreviation-definition pairs are highly related, they can be extracted from these related words. Our method involves three steps. Firstly we utilize external statistical features to extract candidate abbreviation-definition pairs from anchor texts. Secondly we extract internal features from candidate pairs and adopt Conditional Random Fields (CRFs) to compute a score for each candidate pair. Finally we combine external and internal features to generate the final pairs. Experimental results show that this method can accurately extract Chinese abbreviation-definition pairs from anchor texts and combining both external and internal features is effective for extracting abbreviation-definition pairs. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Extracting Chinese-English phrase translation pairs based on tree-tree alignment

    Publication Year: 2011 , Page(s): 1492 - 1496
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (111 KB)  

    Phrase translation pairs are very useful for bilingual lexicography, machine translation system, cross-lingual information retrieval and many applications in natural language processing. Linguistics knowledge in lexicon and statistics information in corpus can be used for extracting phrase translation pairs. In this paper, we propose a new method to extract phrase translation pairs based on aligning Chinese parsing tree and English parsing tree in a bilingual sentence pair. Experimental results indicate that the extracted phrase translation pairs achieve 59.10% at accuracy, when the new method is applied. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fast text categorization based on collaborative work in the semantic and class spaces

    Publication Year: 2011 , Page(s): 1497 - 1502
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1000 KB) |  | HTML iconHTML  

    The blooming of the Internet information has made fast text categorization very essential. Generally, in order to accelerate the classification process, the classifier needs to be simplified as much as possible; however, the accuracy might descend drastically in that case, This paper proposes a novel approach to achieve a suitable tradeoff between the speed and accuracy. With category information fusion and basis orthogonality non-negative matrix factorization, the documents can be mapped from the term space to a semantic or class s-pace, and a simple and fast classification method in the class space is proposed. Furthermore a criterion for re-classifying in the semantic space is discussed. Finally, the collaborative work framework in the semantic and class spaces is implemented. Experiments in two benchmarks are presented, and the results are encouraging. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Incremental learning on background net to capture changing personal reading preference

    Publication Year: 2011 , Page(s): 1503 - 1508
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (185 KB) |  | HTML iconHTML  

    This article proposes a novel approach for capturing user's personal preference of reading by a long-term knowledge background accumulated through incremental learning on user's favorite articles, to better serve personal article selection. User's knowledge background is represented as weighted undirected graph called background net that captures the contextual association of words appeared in the articles recommended. With a background net of user constructed, the understanding of a word is personalized to a fuzzy set based on contextual association of the given word to other words involved in the user's background net. Similarity and acceptance measures are defined to evaluate candidate article through associate reasoning on background net. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Intelligence thresholding for degraded text-photo document images

    Publication Year: 2011 , Page(s): 1509 - 1514
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1032 KB) |  | HTML iconHTML  

    The conversion of the content from paper books into digital form is captured by digital cameras or scanners. However, after the conversion, the illumination of the captured document images is often unevenly distributed. Conventional thresholding methods cannot threshold these kind documents, usually full text images, properly. If the degraded document image includes both text and photo, these methods produce unsatisfactory binarizaion results. This paper presents an efficient and effective intelligent thresholding method for degraded text-photo document images, including: gray-level region cutting is proposed to segment the gray-level document image into several regions intelligently; each region is thresholded by using region thresholding; the gray-level document image is converted into a binary image. Experimental results show that the performance of the proposed method is better than other available thresholding methods in visual measurement. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multi-feature representation for Web-based English-Chinese OOV term translation

    Publication Year: 2011 , Page(s): 1515 - 1519
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (159 KB) |  | HTML iconHTML  

    This paper focuses on the Web-based English-Chinese Out-of-Vocabulary (OOV) term translation pattern, and emphasizes particularly on the selection strategy based on the multi-feature representation for translation evaluation. Three kinds of feature, local feature, global feature and Boolean feature, are extracted from translation candidates based on the fusion strategy of multi-features. By utilizing the CoNLL 2003 corpus for the English Named Entity Recognition (NER) task, the related experiments based on such a standard data source show the promising results. The established multi-feature representation mechanism for English-Chinese OOV term translation model can “filter” the most possible translation candidate with better ability. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Real-time camera anomaly detection for real-world video surveillance

    Publication Year: 2011 , Page(s): 1520 - 1525
    Cited by:  Papers (5)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (927 KB) |  | HTML iconHTML  

    This paper proposes an automatic event detection technique for camera anomaly by image analysis, in order to confirm good image quality and correct field of view of surveillance videos. The technique first extracts reduced-reference features from multiple regions in the surveillance image, and then detects anomaly events by analyzing variation of features when image quality decreases and field of view changes. Event detection is achieved by statistically calculating accumulated variations along temporal domain. False alarms occurred due to noise are further reduced by an online Kalman filter that can recursively smooth the features. Experiments are conducted on a set of recorded videos simulating various challenging situations. Compared with an existing method, experimental results demonstrate that our method has high precision and low false alarm rate with low time complexity. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Similarity measures between intervals of linguistic 2-tuples and the intervals of linguistic 2-tuples weighted average operator

    Publication Year: 2011 , Page(s): 1526 - 1531
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (180 KB) |  | HTML iconHTML  

    In this paper, we present a similarity measure between intervals of linguistic 2-tuples and present the intervals of linguistic 2-tuples weighted average operator. The proposed the proposed similarity measure between intervals of linguistic 2-tuples and the proposed intervals of linguistic 2-tuples weighted average operator can be used for fuzzy group decision making based on the linguistic 2-tuples representation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Text classification using word sequence kernel methods

    Publication Year: 2011 , Page(s): 1532 - 1537
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (564 KB) |  | HTML iconHTML  

    This paper presents a comparison study of two sequence kernels for text classification, namely, all common subsequences and sequence kernel. We consider some variations of the two kernels - kernels based on individual features, linear combination of individual kernels and kernels with a factored representation of features - and evaluate them in text classification by employing them as similarity functions in a support vector machine. A sentence is represented as a sequence of words along with their lemma and part-of-speech tags. Experiments show that sequence kernel has a clear advantage over all common subsequences. Since the main difference between the two kernels lies in the fact that the frequency of words (objects) is considered in sequence kernel but not in all common subsequences, we conclude that the frequency of words is an important factor in the successful application of kernels to text classification. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Text detection in images based on Multiple Kernel Learning

    Publication Year: 2011 , Page(s): 1538 - 1543
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1042 KB) |  | HTML iconHTML  

    Detecting text accurately is an essential requirement for text recognition. In this paper, we propose a method to automatically detect text information in images. We firstly find the candidates of text regions based on the analysis of connected components and extract textural features in these candidate regions. We apply Multiple Kernel Learning to train a classifier with an optimal combination of kernels. The classifier can be used to distinguish text from icons which might be included in region candidates. Our method has been successfully implemented in detecting text from the interface images of mobile phones. According to the experimental results, our method outperforms several typical SVM based methods. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Thai word segmentation for visualization of Thai Web sites

    Publication Year: 2011 , Page(s): 1544 - 1549
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (265 KB) |  | HTML iconHTML  

    Information overload is a problem in the Information Age and Information visualization is an approach to provide an overview of the content of a web site. Tag cloud is one of the ways to represent information as an image of a group of words. However, there are limitations on tag cloud generation, and one of them is due to the characteristics for the language. In order to extract tags or words for tag cloud, word segmentation is required. This paper proposes a Thai word segmentation approach for the visualization of Thai Web sites. The proposed Thai word segmentation technique is based on the longest matching technique together with a refined corpus. The results of Thai word segmentation are compatible with the results from previous BEST's contests in Thailand. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Video-based intelligent vehicle contextual information extraction for night conditions

    Publication Year: 2011 , Page(s): 1550 - 1554
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (185 KB) |  | HTML iconHTML  

    Advanced warning system for vehicles is a critical issue in recent years for automobiles, especially when the number of vehicles is growing rapidly world wide. The cost down of general cameras makes it feasible to have an intelligent system of visual-based event detection in front for forward collision avoidance and mitigation. When driving at nighttime, vehicles in front are generally visible by their taillights. Therefore, in this paper, a computational system, which is referred to as the dynamic visual system, is proposed to detect and analyze the taillights of the vehicles in front in spatiotemporal domain, and then extract corresponding contextual information. Predefined critical contextual information of nearby vehicles can be used for driver-assistance systems to convey a warning. Experiment from extensive dataset shows that our proposed system can effectively extract critical contextual information under different lighting and traffic conditions, and thus prove its feasibility in real-world environments. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.