Joint Matrix Factorization Research Articles

MC 2 : Unsupervised Multiple Social Network Alignment

Social network alignment, identifying social accounts of the same individual across different social networks, shows fundamental importance in a wide spectrum of applications, such as link prediction and information diffusion. Individuals more often than not join in multiple social networks, and it is in fact much too expensive or even impossible to acquiring supervision for guiding the alignment. To the best of our knowledge, few method in the literature can align multiple social networks without supervision. In this article, we propose to study the problem of unsupervised multiple social network alignment. To address this problem, we propose a novel unsupervised model of joint Matrix factorization with a diagonal Cone under orthogonal Constraint, referred to as MC 2 . Its core idea is to embed and align multiple social networks in the common subspace via an unsupervised approach. Specifically, in MC 2 model, we first design a matrix optimization to infer the common subspace from different social networks. To address the nonconvex optimization, we then design an efficient alternating algorithm by leveraging its inherent functional property. Through extensive experiments on real-world datasets, we demonstrate that the proposed MC 2 model significantly outperforms the state-of-the-art methods.

A joint matrix factorization and clustering scheme for irregular time series data

Key Performance Indicator (KPI) clustering plays an important role in Artificial Intelligence for IT Operations (AIOps) when the number of KPIs is large. This approach can effectively reduce the overhead by dividing KPIs into several classes, then applying the same anomaly detection or prediction model to all KPIs in a class. However, KPI sampling strategies vary depending on the environment in question, which leads to the production of irregular KPIs. Few existing works have considered the clustering of KPIs with irregular sampling. Matrix factorization (MF) is widely applied in low-rank data recovery and can be used to align and fill irregular KPIs. However, the clustering performance after recovering and filling by MF remains unknown. These two problems interact with each other and should therefore be solved together. Accordingly, we formulate the joint MF and clustering problem for irregular KPIs and design an iterative clustering scheme based on MF. This iterative clustering scheme comprises alignment and pre-filling, the loop of clustering, and subclass filling by MF, and can work with two pre-filling methods. Extensive experiments on two real-world datasets show that the iterative clustering scheme can obtain higher normalized mutual information (NMI) than non-iterative clustering, while also consuming less computational time than Dynamic Time Warping (DTW). The two kinds of pre-filling methods each have their advantages on different datasets.

Multivariate Transfer Passenger Flow Forecasting with Data Imputation by Joint Deep Learning and Matrix Factorization

Accurate forecasting of the future transfer passenger flow from historical data is essential for helping travelers to adjust their trips, optimal resource allocation and alleviating traffic congestion. However, current studies have mainly emphasized predicting traffic parameters for a single type of transport, while lacking research into transfer passenger flow influenced by multiple factors across different transport modes. Additionally, efficient traffic prediction relies on high-quality traffic data, yet data loss issues are inevitable but often ignored. To fill these gaps, we present for the first time a reliable joint long short-term memory with matrix factorization deep learning model (i.e., Joint-IF) for accurate imputation and forecasting of transfer passenger flow between metro and bus. This hybrid Joint-IF model uses a repair-before-prediction strategy to deliver the final high-quality outputs. In particular, we simulate a variety of missing combinations under the natural conditions and apply a low-rank matrix factorization to infer those lost values. In addition, we investigate the effects of crucial parameters and spatiotemporal features on transfer flow prediction. To validate the effectiveness of Joint-IF, a large series of experiments are carried out for models’ comparison and validation on the real-world transfer passenger flow dataset of the Shenzhen public transport system, and the results show that the proposed Joint-IF performs better for both imputation and forecasting of transfer passenger flow relative to the baseline models in terms of accuracy and stability.

Cooperative driver pathways discovery by multiplex network embedding.

Cooperative driver pathways discovery helps researchers to study the pathogenesis of cancer. However, most discovery methods mainly focus on genomics data, and neglect the known pathway information and other related multi-omics data; thus they cannot faithfully decipher the carcinogenic process. We propose CDPMiner (Cooperative Driver Pathways Miner) to discover cooperative driver pathways by multiplex network embedding, which can jointly model relational and attribute information of multi-type molecules. CDPMiner first uses the pathway topology to quantify the weight of genes in different pathways, and optimizes the relations between genes and pathways. Then it constructs an attributed multiplex network consisting of micro RNAs, long noncoding RNAs, genes and pathways, embeds the network through deep joint matrix factorization to mine more essential information for pathway-level analysis and reconstructs the pathway interaction network. Finally, CDPMiner leverages the reconstructed network and mutation data to define the driver weight between pathways to discover cooperative driver pathways. Experimental results on Breast invasive carcinoma and Stomach adenocarcinoma datasets show that CDPMiner can effectively fuse multi-omics data to discover more driver pathways, which indeed cooperatively trigger cancers and are valuable for carcinogenesis analysis. Ablation study justifies CDPMiner for a more comprehensive analysis of cancer by fusing multi-omics data.

Cross-domain recommendation with Multi-Auxiliary Domains via Consistent and Selective Cluster-Level Knowledge Transfer

Recommender systems are software tools and techniques designed to provide users with useful items suggestions. For most existing recommender systems, data sparsity is a challenging and common problem. The cross-domain recommender systems can mitigate the sparsity problem of data in a target domain by introducing auxiliary domain with relatively dense data. However, there are three challenging issues in the existing cross-domain recommender systems: (1) Most cross-domain recommender systems are designed for a single auxiliary domain, which may lead to insufficient knowledge transfer. However, in practice, there may be multiple auxiliary domains related to the target domain that are available; (2) Due to the distribution divergence between the auxiliary and target domains, inconsistent knowledge may be transferred to the target domain; (3) Because of the diversity among domains, they do not necessarily share exactly the same rating pattern. Moreover, not all cluster-level knowledge of auxiliary domains is beneficial to the target domain, and some cluster-level knowledge may be harmful to the target domain. To deal with the above three issues, we propose a Cross-Domain Recommendation with Multi-Auxiliary Domains via Consistent and Selective Cluster-Level Knowledge Transfer, called MD-CSKT. First, rating matrices from multiple auxiliary domains are decomposed jointly with a single target rating matrix to transfer richer knowledge from multiple auxiliary domains. Secondly, the Maximum Mean Discrepancy (MMD) regularization constraint is incorporated into the joint decomposition to reduce the distribution discrepancy between each auxiliary domain and the target domain, so as to achieve consistent cluster-level knowledge transfer. Thirdly, selective cluster-level knowledge transfer is realized by adaptively selecting of cluster-level rating knowledge from multiple auxiliary domains in the joint matrix factorization. The experimental results on six real-world datasets in three categories demonstrate that the presented MD-CSKT method substantially outperforms nine comparison methods and achieves the improvement of recommendation accuracy in the target domain. Compared with the state-of-the-art comparison methods, the overall performance of the MD-CSKT method is improved by more than 25%.

One step multi-view spectral clustering via joint adaptive graph learning and matrix factorization

Multi-view clustering based on graph learning has attracted extensive attention due to its simplicity and efficiency in recent years. However, there are still some issues in most of the existing graph-based multi-view clustering methods. First, most of those methods require post-processing such as K-means or spectral rotation to get the final discrete clustering result. Second, graph-based clustering methods perform clustering on a fixed input similarity graph, which could induce bad clustering results if the initial graph is with low quality. Third, these methods have high computation cost, which hinders them for dealing with large-scale data. In order to solve these problems, we propose a multi-view spectral clustering method via joint Adaptive Graph Learning and Matrix Factorization (AGLMF). In this method, to reduce computational cost, we adopt the anchor-based strategy to construct the input similarity graphs. Then, we use the l1-norm to learn a high quality similarity graph adaptively from original similarity graphs which can make the final graph more robust than original ones. In addition, AGLMF uses symmetric non-negative matrix factorization to learn the final clustering indicators which can show the final consistent clustering result directly. Finally, experimental results on multiple multi-view datasets validate the effectiveness of the proposed algorithm when compared with previous multi-view spectral clustering algorithms. The demo code of this work is publicly available at https://github.com/theywq/AGLMF.git.

Towards machine-learning-driven effective mashup recommendations from big data in mobile networks and the Internet-of-Things

A large number of Web APIs have been released as services in mobile communications, but the service provided by a single Web API is usually limited. To enrich the services in mobile communications, developers have combined Web APIs and developed a new service, which is known as a mashup. The emergence of mashups greatly increases the number of services in mobile communications, especially in mobile networks and the Internet-of-Things (IoT), and has encouraged companies and individuals to develop even more mashups, which has led to the dramatic increase in the number of mashups. Such a trend brings with it big data, such as the massive text data from the mashups themselves and continually-generated usage data. Thus, the question of how to determine the most suitable mashups from big data has become a challenging problem. In this paper, we propose a mashup recommendation framework from big data in mobile networks and the IoT. The proposed framework is driven by machine learning techniques, including neural embedding, clustering, and matrix factorization. We employ neural embedding to learn the distributed representation of mashups and propose to use cluster analysis to learn the relationship among the mashups. We also develop a novel Joint Matrix Factorization (JMF) model to complete the mashup recommendation task, where we design a new objective function and an optimization algorithm. We then crawl through a real-world large mashup dataset and perform experiments. The experimental results demonstrate that our framework achieves high accuracy in mashup recommendation and performs better than all compared baselines.

Machine Learning-Driven APPs Recommendation for Energy Optimization in Green Communication and Networking for Connected and Autonomous Vehicles

With the rapid development of connected and autonomous vehicles (CAVs), a large number of mobile and edge applications (APPs) have been developed and deployed through green communication and networking technology. The problem of high energy consumption during APPs usage becomes serious and in this paper, we propose to optimize energy usage through effective APPs recommendation. Traditional recommendation methods have been developed for years, such as collaborative filtering and latent factor models. But those methods are not designed for APPs recommendation and only focus on the use of historical records. We find that there are hidden relationships in the content and context of APPs in green communication and networking. In this paper, we develop a holistic APPs recommendation framework for CAVs in green communication and networking. The developed framework is driven by machine learning, where we propose two joint matrix factorization models and hidden relationship mining method. The machine learning-driven models can leverage the neglected information and learn latent features in APPs recommendation for CAVs. We used a real-word mobile and edge APPs dataset, performed sufficient experiments and compared our framework with well-known methods. Experimental results show that our framework produces the best performance in all test cases.

Unsupervised Multiple Change Detection for Multispectral Images Based on AMMF and SpatioSpectral Channel Augmentation

Due to the difficulty and time-consuming of labeling ground truth map in practical situations, unsupervised multiple change detection (MCD) for multispectral images (MSIs) have attracted much attention in recent years. One possible strategy to obtain multiple changes is to assign labels to the binary change result. However, some methods are difficult to obtain the accurate binary result because of the complexity of backgrounds; moreover, assigning labels is also a challenge owing to the limitation of the number of spectral channels in MSIs. Therefore, we propose a novel unsupervised MCD framework based on auto-updating multitemporal matrix factorization (AMMF) and spatiospectral channel augmentation (SSCA). In AMMF, the accurate binary change result can be detected based on joint matrix factorization, during which the distribution and subspace information of each temporal image are regularized to encode the spatiotemporal correlation. In SSCA, some novel augmentation strategies are introduced to increase the number of channels in MSIs to form the normalized high-dimensional maps for each temporal image based on nonlinear operations and convolutional sparse analysis, respectively. MCD can be achieved by integrating the binary change result and directional information that can be calculated by high-dimensional maps of different temporal images. Experiments are conducted on two real MSIs, indicating that the proposed framework performs well in detecting multiple changes.

Collaborative APIs recommendation for Artificial Intelligence of Things with information fusion

With the rapid development of Artificial Intelligence of Things (AIoT), many applications are developed and deployed, especially mobile applications and edge applications. Many softwares are developed to support such diverse applications. To facilitate the development of softwares for AIoT, developers and programmers usually rely on the employment of the mature application programming interfaces (APIs). However, it is difficult for developers to select the most suitable APIs to finish the development, decreasing the development efficiency and delaying the development schedule. To solve these problems, in this paper, we propose a collaborative framework for APIs recommendation for AIoT, and also propose a joint matrix factorization technique for information fusion to fully use different types of information in AIoT. The built framework uses the invocation records among users and APIs during the development. Using collaborative technologies, we yield the similarity relationships among users and among APIs and build three novel APIs recommendation models. We collected a real-world dataset and performed sufficient experiments. The experimental results demonstrate that our models produce superior recommendation accuracy.

Self-representation and matrix factorization based multi-view clustering

Although the promising clustering performance, existing self-representation based multi-view subspace clustering methods directly minimize the divergence between affinity matrices to learn the consensus affinity matrix. This does not make sense for multi-view clustering due to the facts that multi-view data are often a collection of distinct attributes of the objects, and each view includes some contents of the objects that other views do not. Thus, the learned affinity representation is sub-optimal and cannot well characterize the cluster structure. To handle this problem, drawing the inspiration from matrix factorization, which lends embedding representation to clustering interpretation, we propose a novel multi-view subspace clustering method. Our method learns affinity representation between data by joint self-representation and matrix factorization with weighted tensor Schatten p-norm constraint. Moreover, auto-weighted strategy is introduced to adaptively characterize the difference between singular values to improve the stableness of the algorithm. To further characterize class-specificity distribution, which well encodes cluster structure, we employ the ℓ1,2-norm regularization on affinity representation. Experimental results on several data sets indicate that our method outperforms state-of-the-art self-representation based multi-view subspace clustering methods.

Missing data imputation for traffic congestion data based on joint matrix factorization

In reality, the missing of some traffic data is inevitable due to some unexpected errors, which not only affects traffic management but also hinders the development of traffic data research. In this paper, we propose a novel Imputation Model for traffic Congestion data, CIM for short, based on joint matrix factorization. CIM jointly models the characteristics of traffic congestion patterns, including periodicity, road similarity and temporal coherence to estimate the missing congestion values. In particular, we first construct an order-3 tensor based on the traffic congestion data. Then, we model the periodicity and road similarity via joint matrix factorization by exploiting the spatial and temporal information. Finally, we incorporate the local constraints into the process of matrix factorization to ensure the temporal coherence. Experimental results on a real traffic dataset indicate that modeling the three features of congestion patterns simultaneously is effective and CIM outperforms the baselines for the task of missing traffic data imputation.

IsoDA: Isoform-Disease Association Prediction by Multiomics Data Fusion.

A gene can be spliced into different isoforms by alternative splicing, which contributes to the functional diversity of protein species. Computational prediction of gene-disease associations (GDAs) has been studied for decades. However, the process of identifying the isoform-disease associations (IDAs) at a large scale is rarely explored, which can decipher the pathology at a more granular level. The main bottleneck is the lack of IDAs in current databases and the multilevel omics data fusion. To bridge this gap, we propose a computational approach called Isoform-Disease Association prediction by multiomics data fusion (IsoDA) to predict IDAs. Based on the relationship between a gene and its spliced isoforms, IsoDA first introduces a dispatch and aggregation term to dispatch gene-disease associations to individual isoforms, and reversely aggregate these dispatched associations to their hosting genes. At the same time, it fuses the genome, transcriptome, and proteome data by joint matrix factorization to improve the prediction of IDAs. Experimental results show that IsoDA significantly outperforms the related state-of-the-art methods at both the gene level and isoform level. A case study further shows that IsoDA credibly identifies three isoforms spliced from apolipoprotein E, which have individual associations with Alzheimer's disease, and two isoforms spliced from vascular endothelial growth factor A, which have different associations with coronary heart disease. The codes of IsoDA are available at http://mlda.swu.edu.cn/codes.php?name=IsoDA.

Deep sparse autoencoder prediction model based on adversarial learning for cross-domain recommendations

Online recommender systems generally suffer from severe data sparsity problems, and this are particularly prevalent in newly launched systems that do not have sufficient amounts of data. Cross-domain recommendations can provide us with some new ideas for assisting with user recommendations in sparse target domains by transferring knowledge from a source domain with rich data. In this paper, a deep sparse autoencoder prediction model based on adversarial learning for cross-domain recommendations (DSAP-AL) is proposed to improve the accuracy of rating predictions in similar cross-domain recommender systems. Specifically, joint matrix factorization and adversarial network learning models are adopted to integrate and align user and item latent factor spaces in a unified pattern. Then, a deep sparse autoencoder is represented and modeled by transferring the latent factors and interlayer weights. Furthermore, a domain factor adaptation algorithm is proposed to capture robust user and item factors, and the learned regularization constraints are added to the objective function, thereby alleviating the data sparsity issue. Experimental results on four real-world datasets demonstrate that, even without overlapping entities (users or items) in the source and target domains, the proposed DSAP-AL method achieves competitive performance relative to other state-of-the-art individual and cross domain approaches. Moreover, the DSAP-AL model is not only effective for scenarios with sparse data but also robust for noise-containing recommendations.

Preference discovery from wireless social media data in APIs recommendation

In recent years, with the development of software development, a large number of developers develop software by invoking API. With the increasing number of APIs, how to accurately recommend the APIs to developers has become a urgently necessary task. In this paper, we discover that there is a relationship between the user and the API, and use such relationships and collaborative learning techniques to finish APIs recommendation. We propose a holistic framework that contains three models. In the models, we design a joint matrix factorization technique and try to discover the preference among APIs invocation process. In natural language processing, word embedding is widely used. In our models, we use doc2vec to turn the representation of users and APIs into vector representation and calculate the similarity separately to generate the relationships. Besides the two modes in users side and APIs side, we also propose an ensemble model fully leveraging the preference mined from both users side and APIs side. We conduct the experiments on a real-world dataset and the experimental results show that our models perform better than all compared methods.

Personalized APIs Recommendation With Cognitive Knowledge Mining for Industrial Systems

With the prevalence of web techniques and Internet-of-Things networks, an increasing number of developers build software by invoking existing application programming interfaces (APIs), especially in industrial systems. As the number of existing APIs in industrial systems is large, it is critical to recommend suitable APIs from big APIs data to developers in industrial software development. There have been some approaches proposed for APIs recommendation, but the existing approaches focus on the utilization of historical invocation records but ignore the exploitation of other information in the development process. We find that this ignored information can be mined as cognitive knowledge to learn the behavior rules of developers. In this article, we propose a holistic personalized recommendation framework that contains two individual models and one ensemble model, which are based on joint matrix factorization and cognitive knowledge mining. In the two individual models, we study the hidden relationships among users, which are mined from the APIs following records. We also study the hidden relationships among APIs, which are mined from the content information. We also propose an ensemble model. We crawled a large real-word dataset and conducted sufficient experiments, and compared our framework with well-known existing methods. The experimental results demonstrate that our framework achieves the best performance.

SoftRec: Multi-Relationship Fused Software Developer Recommendation

Collaboration efficiency is of primary importance in software development. It is widely recognized that choosing suitable developers is an efficient and effective practice for improving the efficiency of software development and collaboration. Recommending suitable developers is complex and time-consuming due to the difficulty of learning developers’ expertise and willingness. Existing works focus on learning developers’ expertise and interactions from their explicit historical information and matching them to specific task. However, such procedures may suffer low accuracy because they ignore implicit information, such as (1) developer–developer collaboration relationships, (2) developer–task implicit interaction relationships, and (3) task–task association relationships, etc. To that end, this paper proposes a multi-relationship fused approach for software developer recommendation (termed SoftRec). First, in addition to explicit developer–task interactions, it considers multivariate implicit relationships, including the three types mentioned above. Second, it integrates these relationships based on joint matrix factorization and generates forecast results upon the architecture of deep neural network. Furthermore, we propose a fast update method to address the cold start issue by making online recommendations for new developers and new tasks. Extensive experiments are conducted on two real-world datasets, and a user study is conducted in a well-known software company. The results demonstrate that SoftRec outperforms four state-of-the-art works.

Joint and individual matrix factorization hashing for large-scale cross-modal retrieval

Multimodal hashing methods have gained considerable attention in recent years due to their effectiveness and efficiency for cross-modal similarity searches. Existing multimodal hashing methods either learn unified hash codes for different modalities or learn individual hash codes for each modality and then explore cross-correlations between them. Generally, learning unified hash codes tends to preserve the shared properties of multimodal data and learning individual hash codes tends to preserve the specific properties of each modality. There remains a crucial bottleneck regarding how to learn hash codes that simultaneously preserve the shared properties and specific properties of multimodal data. Therefore, we present a joint and individual matrix factorization hashing (JIMFH) method, which not only learns unified hash codes for multimodal data to preserve their common properties but also learns individual hash codes for each modality to retain its specific properties. The proposed JIMFH learns unified hash codes by joint matrix factorization, which jointly factorizes all modalities into a shared latent semantic space. In addition, JIMFH learns individual hash codes by individual matrix factorization, which separately factorizes each modality into a modal-specific latent semantic space. Finally, unified hash codes and individual hash codes are combined to obtain the final hash codes. In this way, hash codes learned by JIMFH can preserve both the shared properties and specific properties of multimodal data, and therefore the retrieval performance is enhanced. Comprehensive experiments show that the proposed JIMFH performs much better than many state-of-the-art methods on cross-modal retrieval applications.

Joint Matrix Factorization: A Novel Approach for Recommender System

Collaborative filtering (CF) is the most classical method for recommender system, but it is usually suffered from limited performance by the sparseness of user-to-item rating data. Recently, due to the powerful learning feature representation ability, deep learning components are used to leverage auxiliary information to assist recommendation. However, most existing models based on deep learning are incomplete so that merely extracting the item latent representation and ignoring the user parts. Besides, different data are not chosen from current models. This paper proposes a novel probability framework, named as joint matrix factorization (JMF). There are three components in JMF. Firstly, the modified multilayer crossing version of the factorization machine (MFM) is designed to extract the user latent factors based on user behavior information. Moreover, MFM is a general method which can be used to accomplish many tasks in terms of machine learning. Secondly, a modification of Long Short-Term Memory (LSTM), named as bidirectional LSTM (BLSTM), is used to extract the item latent factors of a document sequence from both front and back directions. Finally, we tightly integrate BLSTM and MFM into probabilistic matrix factorization (PMF) to form JMF. Compared with the classical matrix factorization and other integration models, JMF extracts document data as well as user behavioral data as item vectors and user vectors. Extensive experiments on five real-world datasets show the proposed model has better performance compared with the state-of-the-art recommendation methods.

Knowledge Fusion via Joint Tensor and Matrix Factorization

We consider the task of knowledge fusion, an important aspect of cognitive intelligence, with the goal of combining part-of knowledge drawn from different sources. For this, entities and relations are cast into matrix-based representations. Unlike previous work on relation prediction, we consider the challenging setting of graphs with large amounts of completely separate connected components and no overlap between the training and test set entities. In order to address these challenges, we propose a novel cognitively inspired factorization method that jointly factorizes a subject–predicate–object tensor via RESCAL and a similarity matrix via matrix factorization. Our experimental results show that our method significantly outperforms several strong baseline models, including RESCAL and several TransE-style models. The proposed joint factorization of a subject–predicate–object tensor while applying matrix factorization to a similarity matrix obtains substantially higher average accuracy rates than previous approaches. This shows that it can successfully address the challenge of knowledge fusion of disconnected data.