Vietnam Journal of Computer Science

Nowadays, teaching and learning activities in a course are greatly supported by information technologies. Forums are among information technologies utilized in a course to encourage students to communicate with lecturers more outside a traditional class. Free-styled textual posts in those communications express the problems that the students are facing as well as the interest and activeness of the students with respect to each topic of a course. Exploiting such textual data in a course forum for course-level student prediction is considered in our work. Due to hierarchical structures in course forum texts, we propose a solution in this paper which combines a deep convolutional neural network (CNN) and a loss function to extract the features from textual data in such a manner that more correct recognitions of instances of the minority class which includes students with failure can be supported. In addition, other numeric data are examined and used for the task so that all the students with and without posts can be predicted in the task. Therefore, our work is the first one that defines and solves this prediction task with heterogeneous educational data at the course level as compared to the existing works. In the proposed solution, Random Forests are suggested as an effective ensemble model suitable for our heterogeneous data when many single prediction models which are random trees can be built for many various subspaces with different random features in a supervised learning process. Experimental results in an empirical evaluation on two real datasets show that a heterogeneous combination of textual and numeric data with a Random Forest model can enhance the effectiveness of our solution to the task. The best accuracy and $F$-measure values can be obtained for early predictions of the students with either success or failure. Such better predictions can help both students and lecturers beware of students’ study and support them in time for ultimate success in a course.

The Biswapped-Torus is a recently reported optoelectronic node-symmetrical member of Biswapped-framework family. In this paper,optimized parallel approach is presented for prefix sum computation on $N\times N$ Biswapped-Torus. The proposed parallel algorithm demands total 7$N+6$ electronic and three optical moves on odd network size or 7$N$ electronic and three optical moves on even network size. The algorithmic performance of the suggested parallel algorithm is also compared with the performances of recently reported optimal prefix sum algorithms on $N\times N$ Biswapped-Mesh and $d_h$-dimensional Biswapped Hyper Hexa-cell. Based on the comparative analysis, Biswapped-Torus claims to map prefix sum faster that require fewer communication moves compared to the Grid-based traditional architecture of biswapped family named Biswapped-Mesh. Moreover, the former also has architectural benefit of node-symmetry that leads to advantages such as easy embedding, mapping and designing of routing algorithms. Compared to symmetrical counter-part of biswapped family named Biswapped-Hyper Hexa-Cell, Biswapped-Torus is cost-efficient, but requires comparatively more communication moves for mapping prefix sum.

Data structures and especially the relationship among the data entities have changed in the last couple of years. The network-like graph representations of data-model are becoming more and more common nowadays, since they are more suitable to depict these, than the well-established relational data-model. The graphs can describe large and complex networks — like social networks — but also capable of storing rich information about complex data. This was mostly of relational data-model trait before. This also can be achieved with the use of the knowledge representation tool called “hypergraphs”. To utilize the possibilities of this model, we need a practical way to store and process hypergraphs. In this paper, we propose a way by which we can store hypergraphs model in the SAP HANA in-memory database system which has a “Graph Core” engine besides the relational data model. Graph Core has many graph algorithms by default however it is not capable to store or to work with hypergraphs neither are any of these algorithms specifically tailored for hypergraphs either. Hence in this paper, besides the case study of the two information systems, we also propose pseudo-code level algorithms to accommodate hypergraph semantics to process our IS model.

Fusion techniques with the aim to leverage the discriminative power of different appearance features for person representation have been widely applied in person re-identification. They are performed by concatenating all feature vectors (known as early fusion) or by combining matching scores of different classifiers (known as late fusion). Previous studies have proved that late fusion techniques achieve better results than early fusion ones. However, majority of the studies focus on determining the suitable weighting schemes that can reflect the role of each feature. The determined weights are then integrated in conventional similarity functions, such as Cosine [L. Zheng, S. Wang, L. Tian, F. He, Z. Liu and Q. Tian, Queryadaptive late fusion for image search and person reidentification, in Proc. IEEE Conf. Computer Vision and Pattern Recognition, 2015, pp. 1741–1750]. The contribution of this paper is two-fold. First, a robust person re-identification method by combining the metric learning with late fusion techniques is proposed. The metric learning method Cross-view Quadratic Discriminant Analysis (XQDA) is employed to learn a discriminant low dimensional subspace to minimize the intra-person distance while maximize the inter-person distance. Moreover, product rule-based and sum rule-based late fusion techniques are applied on these distances. Second, concerning feature engineering, the ResNet extraction process has been modified in order to extract local features of different stripes in person images. To show the effectiveness of the proposed method, both single-shot and multi-shot scenarios are considered. Three state-of-the-art features that are Gaussians of Gaussians (GOG), Local Maximal Occurrence (LOMO) and deep-learned features extracted through a Residual network (ResNet) are extracted from person images. The experimental results on three benchmark datasets that are iLIDS-VID, PRID-2011 and VIPeR show that the proposed method $\mathrm{\text{obtains}}+11.86\%,+3.48$% $\mathrm{\text{and}}+2.22$% of improvement over the best results obtained with the single feature. The proposed method that achieves the accuracy of 85.73%, 93.82% and 50.85% at rank-1 for iLIDS-VID, PRID-2011 and VIPeR, respectively, outperforms different SOTA methods including deep learning ones. Source code is publicly available to facilitate the development of person re-ID system.

Restricted Boltzmann machine (RBM) plays an important role in current deep learning techniques, as most of the existing deep networks are based on or related to generative models and image classification. Many applications for RBMs have been developed for a large variety of learning problems. Recent developments have demonstrated the capacity of RBM to be powerful generative models, able to extract useful features from input data or construct deep artificial neural networks. In this work, we propose a learning algorithm to find the optimal model complexity for the RBM by improving the hidden layer (50–750 layers). Then, we compare and analyze the classification performance in depth of regular RBM use RBM () function, classification RBM use stackRBM() function, and Deep Belief Network (DBN) use DBN() function with the different hidden layer. As a result, Stacking RBM and DBN could improve our classification performance compared to regular RBM.

As road accidents are the leading cause of death for young adults all over the world, it is necessary for the police to evaluate the accident circumstances carefully in order to take appropriate prevention measures. The circumstances of an accident vary in their frequency over time and depend on the local conditions at the accident site. An evaluation under geographical and temporal aspects is therefore necessary. On the basis of the time series, we investigate the various accident circumstances, which show interdependencies with each other, and their influence on the number of accidents. Moreover, a multivariate forecasting is used to indicate the future progression of accidents in different geographical regions. Forecast values are determined with a special extension of the ARIMA method. In order to identify geographical regions of interest, we present two different concepts for segmentation of accident data, which allow the adaptation of police measures to local characteristics.