In this paper we introduce a compactness based clustering algorithm. The compactness of a data class is measured by comparing the inter-subset and intra-subset distances. The class compactness of a subset is defined a...
详细信息
In this paper we introduce a compactness based clustering algorithm. The compactness of a data class is measured by comparing the inter-subset and intra-subset distances. The class compactness of a subset is defined as the ratio of the two distances. A subset is called an isolated cluster (or icluster) if its class compactness is greater than 1. All iclusters make a containment tree. We introduce monotonic sequences of iclusters to simplify the structure of the icluster tree, based on which a clustering algorithm is designed. The algorithm has the following advantages: it is effective on data sets with clusters nonlinearly separated, of arbitrary shapes, or of different densities. The effectiveness of the algorithm is demonstrated by experiments.
Chiaroscuro in art is characterized by strong contrasts between light and dark. An object in a certain light condition has a certain chiaroscuro pattern in appearance; and this pattern is invariant to the changes of i...
详细信息
Chiaroscuro in art is characterized by strong contrasts between light and dark. An object in a certain light condition has a certain chiaroscuro pattern in appearance; and this pattern is invariant to the changes of illumination and pose within a certain extent. In this paper we introduce an object tracking method based on chiaroscuro patterns. The chiaroscuro pattern of an image is computed by running a medial axis transform on selected level sets. The medial axes are pruned and decomposed into segments. These segments make a chiaroscuro pattern. The end points of the segments are collected as key points, which are fed to Kalman filter for object tracking. The effectiveness of the algorithm is demonstrated by experiments.
This paper presents a robust and real time method of license plate localization based on level sets. The proposed algorithm consists of three steps: (1) medial axis transformation of selected level sets, (2) identific...
详细信息
This paper presents a robust and real time method of license plate localization based on level sets. The proposed algorithm consists of three steps: (1) medial axis transformation of selected level sets, (2) identification of sub-strokes, and (3) a multi-resolution window-based analysis for plate localization. Experiments show the effectiveness of our algorithm.
This paper presents an entity answer extraction method based on list web table. Firstly, extract table from page using the features of web page table and label, segment the table that includes the potential entity ans...
详细信息
Separation logic is an extension of Hoare logic for reasoning about mutable heap structure. To represent separation logic in the first-order logic, there are several choices to determine what are constants, what are p...
详细信息
Separation logic is an extension of Hoare logic for reasoning about mutable heap structure. To represent separation logic in the first-order logic, there are several choices to determine what are constants, what are predicates and quantifiers, and whether the commands are taken as atomic or composite. This paper shall give a translation of separation logic into a guarded fragment of the first-order logic, such that the translation is faithful, that is, the translation translates a consistent statement (boolean expression, assertion or specification) of separation logic into a consistent formula in the fragment of the first-order logic. By the decidability of the satisfiability problem of the guarded first-order logic, if the commands are taken as atomic in the first-order logic then the guarded first-order logic translated from separation logic is decidable;if the commands are taken as atomic/composite in the first-order logic then the first-order logic translated from separation logic is undecidable.
Cross-domain text categorization targets on adapting the knowledge learnt from a labeled source-domain to an unla-beled target-domain, where the documents from the source and target domains are drawn from different di...
详细信息
The coordination between agent service and Web service is the key factor for intelligent Web service management in the multi-agent based Web service framework. In view of the drawbacks of existing coordination approac...
详细信息
Services computing is an interdisciplinary subject that devotes to bridging the gap between business services and IT services. It is recognized that Requirements Engineering is fundamental in implementing the service ...
详细信息
Minimum Error Rate Training (MERT) as an effective parameters learning algorithm is widely applied in machine translation and system combination area. However, there exists an ambiguity problem in respect to the train...
详细信息
Minimum Error Rate Training (MERT) as an effective parameters learning algorithm is widely applied in machine translation and system combination area. However, there exists an ambiguity problem in respect to the training goal and it is hard for MERT to tackle, that is different parameters may lead to the same minimum error rate in training but greatly different performances in testing. We propose a novel training objective as the unique goal for training towards, namely partial references, and by use of conditional random fields (CRF) to cast the decoding procedure in system combination as a sequence labeling problem. Experiments on Chinese-English translation test sets show that our approach significantly outperforms the MERT-based baselines with less training time.
Parsing plays an important role in semantic role labeling (SRL) because most SRL systems infer semantic relations from I-best parses. Therefore, parsing errors inevitably lead to labeling mistakes. To alleviate this p...
详细信息
暂无评论