Jianqiang Song’s research while affiliated with Institute of Scientific and Technical Information of China and other places

Discriminative dictionary learning (DDL) has attracted significant attention in the field of image classification. To enhance the classification performance, most existing discriminative dictionary learning methods introduce supervision information on the dictionary to project raw training samples into a coefficient subspace. However, the strict constraint on coefficient features may not conducive to the separation of the training samples from different classes for dictionary learning. In this paper, we propose Relaxed Support Vector based Dictionary Learning (RSVDL) for image recognition, which can efficiently learn coefficient features with powerful discrimination and representation capabilities. By constructing a relaxed coefficient subspace that is closely associated with label information, the discriminative of the learned dictionary is also improved. Experimental results on several benchmark datasets show that the proposed RSVDL method is very effective for various image classification tasks. Moreover, the experiments on more challenging datasets further reveal the state-of-art performance of our method by using with the CNN features.

Recently, Zero-Shot Learning (ZSL) has gained great attention due to its significant classification performance for novel unobserved classes. As seen and unseen classes are completely disjoint, the current ZSL methods inevitably suffer from the domain shift problem when transferring the knowledge between the observed and unseen classes. Additionally, most ZSL methods especially those targeting the semantic space may cause the hubness problem due to their use of nearest-neighbor classifiers in high-dimensional space. To tackle these issues, we propose a novel pathway termed Regularized Label Relaxation-based Stacked Autoencoder (RLRSA) to diminish the domain difference between seen and unseen classes by exploiting an effective label space, which has some notable advantages. First, the proposed method establishes the tight relations among the visual representation, semantic information and label space using via the stacked autoencoder, which is beneficial for avoiding the projection domain shift. Second, by incorporating a slack variable matrix into the label space, our RLRSA method has more freedom to fit the test samples whether they come from the observed or unseen classes, resulting in a very robust and discriminative projection. Third, we construct a manifold regularization based on a class compactness graph to further reduce the domain gap between the seen and unseen classes. Finally, the learned projection is utilized to predict the class label of the target sample, thus the hubness issue can be prevented. Extensive experiments conducted on benchmark datasets clearly show that our RLRSA method produces new state-of-the-art results under two standard ZSL settings. For example, the RLRSA obtains the highest average accuracy of 67.82% on five benchmark datasets under the pure ZSL setting. For the generalized ZSL task, the proposed RLRSA is still highly effective, e.g., it achieves the best H result of 58.9% on the AwA2 dataset.

Abstract Zero‐shot learning (ZSL) is to identify target categories without labeled data, in which semantic information is used to transfer knowledge from some seen categories. In the existing Generalized Zero‐Shot Learning (GZSL) methods, domains shift problem always appeared during generating feature stage. In order to solve this problem, a new method to Auto‐Encode the Synthesis Pseudo Features for the GZSL task (AESPF‐GZSL) is proposed in this manuscript. Specifically, the AESPF‐GZSL method trains the generated features under the semantic auto‐encoder framework and exploits attention mechanism to train the generated features again. Then, the generated features are input to the classifier. The proposed method is performed on three benchmark data sets referred as to AWA, CUB and SUN. The experimental results show that the proposed method achieves the state‐of‐the art classifier accuracy both in ZSL and GZSL settings. In ZSL setting, the classification accuracy of our method is superior to the compared algorithms, improved by 0.40% in AWA and 0.30% in SUN, respectively. And in GZSL setting, the classification accuracy of the method is superior to the comparison algorithm 0.41% in Harmonic mean on AWA, and 1.01%, 0.62%, and 1.05% in training data set, testing data set, and harmonic average on SUN.

This work is devoted to establishing a comparatively accurate classification model between symptoms, constitutions, and regimens for traditional Chinese medicine (TCM) constitution analysis to provide preliminary screening and decision support for clinical diagnosis. However, for the analysis of massive distributed medical data in a cloud platform, the traditional data mining methods have the problems of low mining efficiency and large memory consumption, and long tuning time, an association rules method for TCM constitution analysis (ARA-TCM) is proposed that based on FP-growth algorithm and the open-source distributed file system in Hadoop framework (HDFS) to make full use of its powerful parallel processing capability. Firstly, the proposed method was used to explore the association rules between the 9 kinds of TCM constitutions and symptoms, as well as the regimen treatment plans, so as to discover the rules of typical clinical symptoms and treatment rules of different constitutions and to conduct an evidence-based medical evaluation of TCM effects in constitution-related chronic disease health management. Secondly, experiments were applied on a self-built TCM clinical records database with a total of 30,071 entries and it is found that the top three constitutions are mid constitution (42.3%), hot and humid constitution (31.3%), and inherited special constitution (26.2%), respectively. What is more, there are obvious promotions in the precision and recall rate compared with the Apriori algorithm, which indicates that the proposed method is suitable for the classification of TCM constitutions. This work is mainly focused on uncovering the rules of “disease symptoms constitution regimen” in TCM medical records, but tongue image and pulse signal are also very important to TCM constitution analysis. Therefore, this additional information should be considered into further studies to be more in line with the actual clinical needs.

In the edge computing, service placement refers to the process of installing service platforms, databases, and configuration files corresponding to computing tasks on edge service nodes. In order to meet the latency requirements of new types of applications, service placement in edge computing becomes critical. The service placement strategy must be carried out in accordance with the relevant tasks within the program. However, previous research has paid little attention to related tasks within the application. If the service placement strategy does not consider task relevance, the system will frequently switch services and cause serious system overhead. In this paper, we mainly study the problem of service placement in edge computing. At the same time, we considered the issue of network access point selection during data transmission and the dependencies of task execution. We propose a Dynamic Service Placement List Scheduling (DSPLS) algorithm based on dynamic remaining task service time prediction. We conducted relevant simulation experiments, and our algorithm took the least amount of time to complete the task.