Bui Thu Lam, Le Quy Don Technical University
Deputy Editor-In-Chief
Tran Xuan Nam, Le Quy Don Technical University
Advisory Member
Pham The Long, Le Quy Don Technical University
Nguyen Binh, Posts and Telecommunications Institute of Technology
Dinh The Cuong, Directorate of Information Technology
Luong Chi Mai, Vietnam Academy of Science and Technology
Huynh Quyet Thang, Hanoi University of Science and Technology
Nguyen Huu Thanh, Hanoi University of Science and Technology
Vu Duc Thi, Vietnam National University, Hanoi
Nguyen Thanh Thuy, Vietnam National University, Hanoi
Dao Thanh Tinh, Le Quy Don Technical University
Scientific Secretary
Nguyen Van Giang, Le Quy Don Technical University
Table: Issue 7
ID Paper
[Pdf link]
Extracting linguistic summaries based on hedge algebra approach.
Pham Thi Lan, Ho Cam Ha
Linguistic summarization aims to extract knowledge from databases for supporting in making decisions. The previous studies on this field based on the theory of the fuzzy set. This paper proposes an approach based on hedge algebra to the linguistic data summary problem, and it shows that the association rule problem can be viewed as a special case of the data summary problem. We have implemented this approach in an algorithm to generate the data summaries from a database. The proposed algorithm uses a portion of the idea of Apriori and applies on the database equipped a hedge algebra structure for each quantitative attribute.
A robust and effective incomplete data clustering approach based on learning vector quantization.
Vo Thi Ngoc Chau.
In the real world, incomplete data are often encountered anywhere in a data set. That ubiquity of incomplete data makes a clustering task more challenging. Few of the existing works examined both effectiveness and robustness of the incomplete data clustering algorithms. Some of them are not practical due to a lot of parameters in hybrid approaches and/or cannot handle incompleteness in any object at any dimension. Therefore, our work introduces a new algorithm, iVQ\_nps, which is an incomplete data clustering approach based on learning vector quantization and the nearest prototype strategy for handling incomplete data iteratively. Besides, we define the different adaptive linear functions for a learning rate so that our approach can be effective and practical with only one remaining user-specified parameter that is the number of the resulting clusters. Compare to several existing approaches, iVQ\_nps can produce the clusters of good quality and incomplete data approximation via the experiments on both benchmark and synthetic data sets.
Boosting visual attention with interest points and adaptive bilateral filter.
Dao Nam Anh, Pham Quang Huy, Nguyen Quoc Hung, Nguyen Van Nam, Pham Thi Thuy Loan, Le Hoang.
Spotting image regions in high visual attention is deciding to surveillance, visual data summarizing, error concealment and advertising, detecting major visual appearance regions under narrow capability to perform the vast excess of information gathered through visual system. In this paper we introduce a competent algorithm for visual attention detection yielding plan of salient regions in focus on elucidation of the alliance of interest points and saliency to gain better consideration of fundamental visual process. The map of interest points generated from gradients of second order is spatially relevant to the salient regions. In the implementation, the local gradients allowed initial assumption for a map of interest points to be formed through the use of influence relation between pixels. This is then filtered by adaptive bilateral filter to uncover local divergence which leads to draw the map of saliency. Taken together, a boosting visual attention application that addresses to face and fashion images is presented where the salient map grants visual processing tasks like enhancement and de$-$blurring is indeed driven by the deep of spatial intentness for face preferences or fashion favorites.
Key agreement protocols for symmetric cryptography system.
Hoang Van Viet, Bui The Truyen, Tong Minh Duc, Luu Hong Dung
The paper proposes two new key agreement algorithms for symmetric key cryptography system. The most advantage of the two new protocols is the secret key is established requires only a single or two rounds of transportation as origin Diffie - Hellman key exchange protocol. Moreover, the secret key can be authenticated the origin so it can effectively resistant to spoofing attacks. The article also presents the analysis and assessment of the security level of the new proposed protocols, indicating its potential application in practice is very positive.
A multi-feature integration method for clustering web video search results.
Phuc Quang Nguyen, Tien Do, Thanh Duc Ngo, Duy-Dinh Le, Tu-Anh Hoang Nguyen.
To search videos, users usually use online video search systems. However, the returned search results of these systems are presented as a flat list with many videos of different categories mixed together, and as a result, users find it difficult to locate video clips of interest. Therefore, clustering web video search results is necessary in order to help users to locate videos of interest in quickly. This paper aims to develop our previous researches on clustering web video search results which reported in [1], [2], [3]. The main idea based on analyzing and combining the features extracted from video to evaluate the role of each feature and find the set of appropriate features to improve the clustering quality of web video search results.
Some results from studying the bi-regular matrices for building effective MDS matrices in block cipher AES.
Luong The Dung.
The separation matrix with maximum distance (MDS matrix) is widely used in the cipher and hash function, however the linear transformation by using MDS matrix in many current block ciphers is not effective because the number of occurrences of 1 in matrices is not much and the number of different elements in the matrices is quite large. In this paper, we propose a method to develop the effective MDS matrices based on the bi-regular matrix that maximizes the number of occurrences of 1 and minimizes the number of different elements in MDS matrices, we also present an application of the 8x8 matrix generated by the proposed method to improve the transformation of the diffusion layer of cipher that has been widely used as AES.