Search

Motor imagery electroencephalography is widely used in the brain-computer interface systems. Due to inherent characteristics of electroencephalography signals, accurate and real-time multiclass classification is always challenging. In order to solve this problem, a multiclass posterior probability solution for twin SVM is proposed by the ranking continuous output and pairwise coupling in this paper. First, two-class posterior probability model is constructed to approximate the posterior probability by the ranking continuous output techniques and Platt’s estimating method. Secondly, a solution of multiclass probabilistic outputs for twin SVM is provided by combining every pair of class probabilities according to the method of pairwise coupling. Finally, the proposed method is compared with multiclass SVM and twin SVM via voting, and multiclass posterior probability SVM using different coupling approaches. The efficacy on the classification accuracy and time complexity of the proposed method has been demonstrated by both the UCI benchmark datasets and real world EEG data from BCI Competition IV Dataset 2a, respectively. [ABSTRACT FROM AUTHOR]

Knowledge graph completion aims to solve the problem of incompleteness and sparsity in knowledge graphs. However, the negative sampling strategy in current completion methods samples entities with equal probability and cannot guarantee the quality of negative samples, causing gradient disappearance during the training process. In addition, existing embedding learning methods ignore the diversity and complexity of the entities and relations in a knowledge graph. Therefore, we design a dynamic semantic sampling and correlation embedding completion framework that includes a negative sampling algorithm based on dynamic semantic similarity and a correlation embedding model. The negative-sampling algorithm can gradually explore high-quality negative samples to participate in model training. The embedding model can enrich embeddings by learning the sequential and correlated information of entities and relations in the knowledge graph. Finally, we conducted experiments on two public datasets, and the experimental results proved the performance of our method. [ABSTRACT FROM AUTHOR]

Time-dependent graphs have been well researched in a wealth of fields, such as road network, bioinformatics network. Unlike in static graphs, relations among nodes will change by time in time-dependent graphs, which causes some for such special properties as temporal reachability and node dynamic local structure. And, for different kinds of time-dependent graphs, activity frequency of nodes may be greatly different. These properties should be taken into consideration while embedding nodes in time-dependent graphs into vectors for further research. So, in this work, we study the problem of time-dependent graph embedding and propose a structure similarity based adaptive sampling method, called ATDGEB (A daptive T ime- D ependent G raph E m b edding), which aims to encode different kinds of time-dependent graph nodes into vectors based on node's local structure and their special properties. Specifically, we first design a new method based on node's local structure to compute visit probability between nodes, and then propose an adaptive clustering method for solving the problem that nodes' active frequency is greatly different in different types of time-dependent graph. Meanwhile to get the walk paths as soon as possible, we design a novel walk strategy to get node's walk paths. The sampled nodes in walk process will be stored in bidirectional multi-tree. Once the walk process is finished, we can get node's walk path by reversely travelling​ the multi-tree from leaf nodes in the tree. Sufficient experiments conducted on real datasets demonstrate that our method outperforms the existing embedding methods with respect to node clustering, reachability prediction, and link prediction. [ABSTRACT FROM AUTHOR]