In this paper, we extend the original non-negative matrix factorization (NMF) to kernel NMF (KNMF). The advantages of KNMF over NMF are: 1) it could extract more useful features hidden in the original data through some kernel-induced nonlinear mappings; 2) it can deal with data where only relationships (similarities or dissimilarities) between objects are known; 3) it can process data with negative values by using some specific kernel functions (e.g. Gaussian). Thus, KNMF is more general than NMF. To further improve the performance of KNMF, we also propose the SpKNMF, which performs KNMF on sub-patterns of the original data. The effectiveness of the proposed algorithms is validated by extensive experiments on UCI datasets and the FERET face database.

Non-negative matrix factorization on kernels

Programming is considered a skill to arouse and inspire learner's potential. Learning to program is a complex process that requires students to write grammar and instructions. The structure of a programming language does not cause impose problems to students, the real obstacle is how to apply these learned grammars and present them in a complete and correct program code for problem solving. In this study, a deep learning recommendation system was developed, which includes augmented reality (AR) technology, and learning theory, and is provided for study by students in non-major and also from different learning backgrounds. Those students divided into two groups, the students participating in the experimental group were using the AR system with deep learning recommendation and the students participating in the control group were using the AR system without deep learning recommendation. The results show that students in experimental group perform better than the control group with regards to learning achievement. On the other hand, in the part of computational thinking ability, students using a deep learning recommendation based AR system is significantly better than those using non-deep learning recommendation based AR system. Among the various dimensions of computational thinking, creativity, logical computing, critical thinking, and problem-solving skills are significantly different among the two groups of students. The students in experimental group perform better than the control group with regards to the dimensions of computational thinking, creativity, logical computing, critical thinking, and problem-solving skills.

/pdf/design-and-evaluation-of-a-deep-learning-recommendation-51a2nwcmow.pdf

Design and Evaluation of a Deep Learning Recommendation Based Augmented Reality System for Teaching Programming and Computational Thinking

Clustering is a fundamental research field and plays an important role in data analysis. To better address the relationship between an element and a cluster, a Three-Way clustering method based on an Improved DBSCAN (3W-DBSCAN) algorithm is proposed in this paper. 3W-DBSCAN represents a cluster by a pair of nested sets called lower bound and upper bound respectively. The two bounds classify objects into three status: belong-to, not belong-to and ambiguity. Objects in lower bound certainly belong to the cluster. Objects in upper bound while not in the lower bound are ambiguous because they are in a boundary region and might belong to one or more clusters. Objects beyond the upper bound certainly do not belong to the same cluster. This clustering representation can well explain the clustering result and consist with human cognitive thinking. By improving similarity calculation, improved DBSCAN is presented to obtain initial clustering results, then three-way decision strategies are used to acquire the positive and boundary regions of a cluster. Three benchmarks Accuracy ( A c c ), F-measure ( F 1 ), N M I and ten datasets including three synthetic datasets, three UCI datasets and four shape datasets are used in experiments to evaluate the effectiveness of 3W-DBSCAN. Experimental results suggest that 3W-DBSCAN has a good performance and is effective in clustering.

A three-way clustering method based on an improved DBSCAN algorithm

Due to the high demands of deep neural network (DNN) based applications on computational capability, it is hard for them to be directly run on mobile devices with limited resources Computation offloading technology offers a feasible solution by offloading some computation-intensive tasks of neural network layers to edges or remote clouds that are equipped with sufficient resources However, the offloading process might lead to excessive delays and thus seriously affect the user experience To address this important problem, we first regard the average response time of multi-task parallel scheduling as our optimization goal Next, the problem of computation offloading and task scheduling for DNN-based applications in cloud-edge computing is formulated with a scheme evaluation algorithm Finally, the greedy and genetic algorithms based methods are proposed to solve the problem The extensive experiments are conducted to demonstrate the effectiveness of the proposed methods for scheduling tasks of DNN-based applications in different cloud-edge environments The results show that the proposed methods can obtain the near-optimal scheduling performance, and generate less average response time than traditional scheduling schemes Moreover, the genetic algorithm leads to less average response time than the greedy algorithm, but the genetic algorithm needs more running time

/pdf/computation-offloading-and-task-scheduling-for-dnn-based-qqtqp1zh9p.pdf

Computation Offloading and Task Scheduling for DNN-Based Applications in Cloud-Edge Computing

The quality of learned features by representation learning determines the performance of learning algorithms and the related application tasks (such as high-dimensional data clustering). As a relatively new paradigm for representation learning, Concept Factorization (CF) has attracted a great deal of interests in the areas of machine learning and data mining for over a decade. Lots of effective CF based methods have been proposed based on different perspectives and properties, but note that it still remains not easy to grasp the essential connections and figure out the underlying explanatory factors from exiting studies. In this paper, we therefore survey the recent advances on CF methodologies and the potential benchmarks by categorizing and summarizing the current methods. Specifically, we first re-view the root CF method, and then explore the advancement of CF-based representation learning ranging from shallow to deep/multilayer cases. We also introduce the potential application areas of CF-based methods. Finally, we point out some future directions for studying the CF-based representation learning. Overall, this survey provides an insightful overview of both theoretical basis and current developments in the field of CF, which can also help the interested researchers to understand the current trends of CF and find the most appropriate CF techniques to deal with particular applications.

A Survey on Concept Factorization: From Shallow to Deep Representation Learning.

Recently, deep learning has gained great popularity in the area of recommender systems. Various combinations of deep learning, collaborative recommendation and content-based recommendation have occurred. However, as one of the three most significant recommendation techniques, hybrid recommendation has little cooperation with deep learning. Besides, most current deep hybrid models only incorporate two simple recommendation methods together in post-fusion, leaving massive space for further exploration of better combinations. In this paper, we apply deep learning to hybrid recommendation, proposing a deep hybrid recommendation model DMFL (Deep Metric Factorization Learning). In DMFL, we combine deep learning with improved machine learning models to learn the interaction between users and items from multiple perspectives. Such deep hybrid learning helps to reflect the user preference more comprehensively and strengthen model’s ability of generalization. We also propose a more accurate method of user feature representation, taking both long-term static characteristics and short-term dynamic interest changes of users into consideration. Furthermore, thorough experiments have been conducted on real-world datasets, which strongly proves the effectiveness and efficiency of the proposed model.

An Efficient Hybrid Recommendation Model With Deep Neural Networks

Community detection is an important problem of complex networks analysis and various methods have been proposed to solve it. However, most of the existing methods only use the link information. As a result, the quality of their detected communities is often poor due to the sparse and noisy data existing in link information. Actually, content information of complex networks can also help to improve the quality of community detection. In this paper, we propose a method based on Multi-View Clustering via Robust Nonnegative Matrix Factorization (MVCRNMF). This method can provide a unified framework to combine link and content information for community detection. Its key idea is to build a multi-view robust NMF model with the co-regularized constraint on community indicator matrices of link view and content view. This can make link and content information complement each other during the factorization process of NMF. We devise iterative update rules as the optimization solution to the community detection model and also give the rigorous convergence proof. It is worth noting that MVCRNMF can learn the contribution weights from link and content information adaptively and this helps to save a lot of time on tuning the weight parameters. We conduct comparative experiments on four real-world complex networks. The results demonstrate that MVCRNMF performs better than state-of-the-art methods. Additionally, results of the case study on a co-authorship network also show that MVCRNMF can obtain higher quality communities.

A robust multi-view clustering method for community detection combining link and content information

Nonnegative matrix factorization (NMF) has become a powerful model for community discovery in complex networks Existing NMF-based methods for community discovery often factorize the corresponding adjacent matrix of complex networks to obtain its community indicator matrix However, the adjacent matrix cannot represent the global structure feature of complex networks very well, and this leads to the performance degradation of community discovery Besides, most of existing methods are not robust and scalable enough, so they are not effective to deal with complex networks with noises and large scales Aiming at these problems above, in this paper we propose a method for community discovery using distributed robust NMF with SimRank similarity measure This method selects SimRank measure to construct the feature matrix, which can more accurately represent the global structure feature of complex networks To improve the robustness, we select $$\ell _{2,1}$$
 norm instead of the widely used Frobenius norm to construct its NMF-based community discovery model In addition, to improve the scalability, we implement its key components by using MapReduce distributed computing framework, including computing SimRank feature matrix and iteratively solving the NMF-based model for community discovery We conduct extensive experiments on several typical complex networks The results show that our method has better performance and robustness than other representative NMF-based methods for community discovery Moreover, our method presents good scalability and hence can be used to discover communities in the large-scale complex networks

Improving NMF-based community discovery using distributed robust nonnegative matrix factorization with SimRank similarity measure

SGXPool: Improving the performance of enclave creation in the cloud

Memory copying is one of the most common operations in modern software. Usually, the operation reflects a synchronous (sync) CPU procedure of memory copying, incurring overheads such as cache pollution and CPU stalling, especially in the scenario of bulk copying with large data. To improve this issue, some works based on I/OAT, which is a dedicated and popular hardware copying engine on Intel platform, is proposed but still exists several problems: (1) lacking atomic allocation/revocation at the granularity of I/OAT channel; (2) deficiency of interrupt support and (3) complicated programming interfaces. We propose RAMCI, an asynchronous (async) memory copying mechanism based on Intel I/OAT engine, not only improves the sync overheads, but also overcomes the above three issues through (1) a lock mechanism by using low-level CAS instruction; (2) a lightweight interrupt mechanism for the completion of memory copying, instead of using the polling pattern which consuming large CPU resource and (3) a group of well-defined and abstract interfaces, allowing the programmers to utilize the underlying free I/OAT channels transparently. To support the interfaces, a novel scheduler of the I/OAT channels is introduced. It splits the source copying data into several pieces, and each of them can be allocated with a dedicated I/OAT channel intelligently to transfer the data with parallelism. We evaluate RAMCI and compare it with other memory copying mechanisms in four NUMA scenarios. The experimental results show that RAMCI improves memory copying performance up to 4.68
 $$\times $$
 while achieving almost full ability of parallel computing.

Hai Liu

Papers

An Efficient Hybrid Recommendation Model With Deep Neural Networks

A robust multi-view clustering method for community detection combining link and content information

Improving NMF-based community discovery using distributed robust nonnegative matrix factorization with SimRank similarity measure

SGXPool: Improving the performance of enclave creation in the cloud

RAMCI: a novel asynchronous memory copying mechanism based on I/OAT