With the demand of agile development and management, cloud applications today are moving towards a more fine-grained microservice paradigm, where smaller and simpler functioning parts are combined for providing end-to-end services. In recent years, we have witnessed many research efforts that strive to optimize the performance of cloud computing system in this new era. This paper provides an overview of existing works on recent system performance optimization techniques and classify them based on their design focuses. We also identify open issues and challenges in this important research direction.
As the boom of mobile devices, Android mobile apps play an irreplaceable roles in people’s daily life, which have the characteristics of frequent updates involving in many code commits to meet new requirements. Just-in-Time (JIT) defect prediction aims to identify whether the commit instances will bring defects into the new release of apps and provides immediate feedback to developers, which is more suitable to mobile apps. As the within-app defect prediction needs sufficient historical data to label the commit instances, which is inadequate in practice, one alternative method is to use the cross-project model. In this work, we propose a novel method, called KAL, for cross-project JIT defect prediction task in the context of Android mobile apps. More specifically, KAL first transforms the commit instances into a high-dimensional feature space using kernel-based principal component analysis technique to obtain the representative features. Then, the adversarial learning technique is used to extract the common feature embedding for the model building. We conduct experiments on 14 Android mobile apps and employ four effort-aware indicators for performance evaluation. The results on 182 cross-project pairs demonstrate that our proposed KAL method obtains better performance than 20 comparative methods.
Various kinds of online social media applications such as Twitter and Weibo, have brought a huge volume of short texts. However, mining semantic topics from short texts efficiently is still a challenging problem because of the sparseness of word-occurrence and the diversity of topics. To address the above problems, we propose a novel supervised pseudo-document-based maximum entropy discrimination latent Dirichlet allocation model (PSLDA for short). Specifically, we first assume that short texts are generated from the normal size latent pseudo documents, and the topic distributions are sampled from the pseudo documents. In this way, the model will reduce the sparseness of word-occurrence and the diversity of topics because it implicitly aggregates short texts to longer and higher-level pseudo documents. To make full use of labeled information in training data, we introduce labels into the model, and further propose a supervised topic model to learn the reasonable distribution of topics. Extensive experiments demonstrate that our proposed method achieves better performance compared with some state-of-the-art methods.
It is well known that deep learning depends on a large amount of clean data. Because of high annotation cost, various methods have been devoted to annotating the data automatically. However, a larger number of the noisy labels are generated in the datasets, which is a challenging problem. In this paper, we propose a new method for selecting training data accurately. Specifically, our approach fits a mixture model to the per-sample loss of the raw label and the predicted label, and the mixture model is utilized to dynamically divide the training set into a correctly labeled set, a correctly predicted set, and a wrong set. Then, a network is trained with these sets in the supervised learning manner. Due to the confirmation bias problem, we train the two networks alternately, and each network establishes the data division to teach the other network. When optimizing network parameters, the labels of the samples fuse respectively by the probabilities from the mixture model. Experiments on CIFAR-10, CIFAR-100 and Clothing1M demonstrate that this method is the same or superior to the state-of-the-art methods.
Music is the language of emotions. In recent years, music emotion recognition has attracted widespread attention in the academic and industrial community since it can be widely used in fields like recommendation systems, automatic music composing, psychotherapy, music visualization, and so on. Especially with the rapid development of artificial intelligence, deep learning-based music emotion recognition is gradually becoming mainstream. This paper gives a detailed survey of music emotion recognition. Starting with some preliminary knowledge of music emotion recognition, this paper first introduces some commonly used evaluation metrics. Then a three-part research framework is put forward. Based on this three-part research framework, the knowledge and algorithms involved in each part are introduced with detailed analysis, including some commonly used datasets, emotion models, feature extraction, and emotion recognition algorithms. After that, the challenging problems and development trends of music emotion recognition technology are proposed, and finally, the whole paper is summarized.
Action recognition is an important research topic in video analysis that remains very challenging. Effective recognition relies on learning a good representation of both spatial information (for appearance) and temporal information (for motion). These two kinds of information are highly correlated but have quite different properties, leading to unsatisfying results of both connecting independent models (e.g., CNN-LSTM) and direct unbiased co-modeling (e.g., 3DCNN). Besides, a long-lasting tradition on this task with deep learning models is to just use 8 or 16 consecutive frames as input, making it hard to extract discriminative motion features. In this work, we propose a novel network structure called ResLNet (Deep Residual LSTM network), which can take longer inputs (e.g., of 64 frames) and have convolutions collaborate with LSTM more effectively under the residual structure to learn better spatial-temporal representations than ever without the cost of extra computations with the proposed embedded variable stride convolution. The superiority of this proposal and its ablation study are shown on the three most popular benchmark datasets: Kinetics, HMDB51, and UCF101. The proposed network could be adopted for various features, such as RGB and optical flow. Due to the limitation of the computation power of our experiment equipment and the real-time requirement, the proposed network is tested on the RGB only and shows great performance.
Automatically generating a brief summary for legal-related public opinion news (LPO-news, which contains legal words or phrases) plays an important role in rapid and effective public opinion disposal. For LPO-news, the critical case elements which are significant parts of the summary may be mentioned several times in the reader comments. Consequently, we investigate the task of comment-aware abstractive text summarization for LPO-news, which can generate salient summary by learning pivotal case elements from the reader comments. In this paper, we present a hierarchical comment-aware encoder (HCAE), which contains four components: 1) a traditional sequenceto-sequence framework as our baseline; 2) a selective denoising module to filter the noisy of comments and distinguish the case elements; 3) a merge module by coupling the source article and comments to yield comment-aware context representation; 4) a recoding module to capture the interaction among the source article words conditioned on the comments. Extensive experiments are conducted on a large dataset of legal public opinion news collected from micro-blog, and results show that the proposed model outperforms several existing state-of-the-art baseline models under the ROUGE metrics.
The latest advance in recommendation shows that better user and item representations can be learned via performing graph convolutions on the user-item interaction graph. However, such finding is mostly restricted to the collaborative filtering (CF) scenario, where the interaction contexts are not available. In this work, we extend the advantages of graph convolutions to context-aware recommender system (CARS, which represents a generic type of models that can handle various side information). We propose Graph Convolution Machine (GCM), an end-to-end framework that consists of three components: an encoder, graph convolution (GC) layers, and a decoder. The encoder projects users, items, and contexts into embedding vectors, which are passed to the GC layers that refine user and item embeddings with context-aware graph convolutions on the user-item graph. The decoder digests the refined embeddings to output the prediction score by considering the interactions among user, item, and context embeddings. We conduct experiments on three real-world datasets from Yelp and Amazon, validating the effectiveness of GCM and the benefits of performing graph convolutions for CARS.
Video question answering (Video QA) involves a thorough understanding of video content and question language, as well as the grounding of the textual semantic to the visual content of videos. Thus, to answer the questions more accurately, not only the semantic entity should be associated with certain visual instance in video frames, but also the action or event in the question should be localized to a corresponding temporal slot. It turns out to be a more challenging task that requires the ability of conducting reasoning with correlations between instances along temporal frames. In this paper, we propose an instance-sequence reasoning network for video question answering with instance grounding and temporal localization. In our model, both visual instances and textual representations are firstly embedded into graph nodes, which benefits the integration of intra- and inter-modality. Then, we propose graph causal convolution (GCC) on graph-structured sequence with a large receptive field to capture more causal connections, which is vital for visual grounding and instance-sequence reasoning. Finally, we evaluate our model on TVQA+ dataset, which contains the groundtruth of instance grounding and temporal localization, three other Video QA datasets and three multimodal language processing datasets. Extensive experiments demonstrate the effectiveness and generalization of the proposed method. Specifically, our method outperforms the state-of-the-art methods on these benchmarks.
The hardness of NTRU problem affects heavily on the securities of the cryptosystems based on it. However, we could only estimate the hardness of the specific parameterized NTRU problems from the perspective of actual attacks, and whether there are worst-case to average-case reductions for NTRU problems like other lattice-based problems (e.g., the Ring-LWE problem) is still an open problem. In this paper, we show that for any algebraic number field
Cybersecurity has always been the focus of Internet research. An LDoS attack is an intelligent type of DoS attack, which reduces the quality of network service by periodically sending high-speed but short-pulse attack traffic. Because of its concealment and low average rate, the traditional DoS attack detection methods are challenging to be effective. The existing LDoS attack detection methods generally have the problems of high FPR and FNR. A cloud model-based LDoS attack detection method is proposed, and a classifier based on SVM is used to train and classify the feature parameters. The detection method is verified and tested in the NS2 simulation platform and Test-bed network environment. Compared with the existing research results, the proposed method requires fewer samples, and it has lower FPR and FNR.
Searchable symmetric encryption (SSE) has been introduced for secure outsourcing the encrypted database to cloud storage, while maintaining searchable features. Of various SSE schemes, most of them assume the server is honest but curious, while the server may be trustless in the real world. Considering a malicious server not honestly performing the queries, verifiable SSE (VSSE) schemes are constructed to ensure the verifiability of the search results. However, existing VSSE constructions only focus on single-keyword search or incur heavy computational cost during verification. To address this challenge, we present an efficient VSSE scheme, built on OXT protocol (Cash et al., CRYPTO 2013), for conjunctive keyword queries with sublinear search overhead. The proposed VSSE scheme is based on a privacy-preserving hash-based accumulator, by leveraging a well-established cryptographic primitive, Symmetric Hidden Vector Encryption (SHVE). Our VSSE scheme enables both correctness and completeness verifiability for the result without pairing operations, thus greatly reducing the computational cost in the verification process. Besides, the proposed VSSE scheme can still provide a proof when the search result is empty. Finally, the security analysis and experimental evaluation are given to demonstrate the security and practicality of the proposed scheme.
Genotyping of structural variations considering copy number variations (CNVs) is an infancy and challenging problem. CNVs, a prevalent form of critical genetic variations that cause abnormal copy numbers of large genomic regions in cells, often affect transcription and contribute to a variety of diseases. The characteristics of CNVs often lead to the ambiguity and confusion of existing genotyping features and algorithms, which may cause heterozygous variations to be erroneously genotyped as homozygous variations and seriously affect the accuracy of downstream analysis. As the allelic copy number increases, the error rate of genotyping increases sharply. Some instances with different copy numbers play an auxiliary role in the genotyping classification problem, but some will seriously interfere with the accuracy of the model. Motivated by these, we propose a transfer learning-based method to genotype structural variations accurately considering CNVs. The method first divides the instances with different allelic copy numbers and trains the basic machine learning framework with different genotype datasets. It maximizes the weights of the instances that contribute to classification and minimizes the weights of the instances that hinder correct genotyping. By adjusting the weights of the instances with different allelic copy numbers, the contribution of all the instances to genotyping can be maximized, and the genotyping errors of heterozygote variations caused by CNVs can be minimized. We applied the proposed method to both the simulated and real datasets, and compared it to some popular algorithms including GATK, Facets and Gindel. The experimental results demonstrate that the proposed method outperforms the others in terms of accuracy, stability and efficiency. The source codes have been uploaded at github/TrinaZ/CNVtransfer for academic use only.
The kernel method, especially the kernel-fusion method, is widely used in social networks, computer vision, bioinformatics, and other applications. It deals effectively with nonlinear classification problems, which can map linearly inseparable biological sequence data from low to high-dimensional space for more accurate differentiation, enabling the use of kernel methods to predict the structure and function of sequences. Therefore, the kernel method is significant in the solution of bioinformatics problems. Various kernels applied in bioinformatics are explained clearly, which can help readers to select proper kernels to distinguish tasks. Mass biological sequence data occur in practical applications. Research of the use of machine learning methods to obtain knowledge, and how to explore the structure and function of biological methods for theoretical prediction, have always been emphasized in bioinformatics. The kernel method has gradually become an important learning algorithm that is widely used in gene expression and biological sequence prediction. This review focuses on the requirements of classification tasks of biological sequence data. It studies kernel methods and optimization algorithms, including methods of constructing kernel matrices based on the characteristics of biological sequences and kernel fusion methods existing in a multiple kernel learning framework.
