High-dimensional and incomplete(HDI) matrices are commonly encountered in various big data-related applications for illustrating the complex interactions among numerous entities, like the user-item interactions in a c...High-dimensional and incomplete(HDI) matrices are commonly encountered in various big data-related applications for illustrating the complex interactions among numerous entities, like the user-item interactions in a commercial recommender system or the user-user interactions in a social network services system. The factorization of such an HDI matrix can embed the involved entities into the low-dimensional feature space for acquiring their principal representation, which is a vital task in various application scenes and is often established through the Latent Factor Analysis(LFA). Nevertheless, an HDI matrix can be huge when the corresponding application explodes to involve millions of users, items, or other interactive nodes. In this case, a parallel optimization algorithm is desired for raising the scalability and time efficiency of an LFA model. This paper provides a comprehensive review of the existing parallel optimization algorithms for the LFA model. Specifically, it performs: 1) discussion and summary of these algorithms based on computing architecture and mode, 2) empirical studies of representative models, and3) summary of the current challenges and future directions in this domain. This survey aims to offer an exhaustive review of Parallel Optimization Algorithms for High-Dimensional and Incomplete Matrix Factorization, thereby fostering further research in this field.展开更多
A spacecraft attitude estimation method based on electromagnetic vector sensors(EMVS)array is proposed,which employs the orthogonally constrained parallel factor(PARAFAC)algorithm and makes use of measurements of the ...A spacecraft attitude estimation method based on electromagnetic vector sensors(EMVS)array is proposed,which employs the orthogonally constrained parallel factor(PARAFAC)algorithm and makes use of measurements of the two-dimensional direction-of-arrival(2D-DOA)and polarization angles,aiming to address the issues of incomplete,asynchronous,and inaccurate third-party reference used for attitude estimation in spacecraft docking missions by employing the electromagnetic wave’s three-dimensional(3D)wave structure as a complete third-party reference.Comparative analysis with state-ofthe-art algorithms shows significant improvements in estimation accuracy and computational efficiency with this algorithm.Numerical simulations have verified the effectiveness and superiority of this method.A high-precision,reliable,and cost-effective method for rapid spacecraft attitude estimation is provided in this paper.展开更多
A recommender system(RS)relying on latent factor analysis usually adopts stochastic gradient descent(SGD)as its learning algorithm.However,owing to its serial mechanism,an SGD algorithm suffers from low efficiency and...A recommender system(RS)relying on latent factor analysis usually adopts stochastic gradient descent(SGD)as its learning algorithm.However,owing to its serial mechanism,an SGD algorithm suffers from low efficiency and scalability when handling large-scale industrial problems.Aiming at addressing this issue,this study proposes a momentum-incorporated parallel stochastic gradient descent(MPSGD)algorithm,whose main idea is two-fold:a)implementing parallelization via a novel datasplitting strategy,and b)accelerating convergence rate by integrating momentum effects into its training process.With it,an MPSGD-based latent factor(MLF)model is achieved,which is capable of performing efficient and high-quality recommendations.Experimental results on four high-dimensional and sparse matrices generated by industrial RS indicate that owing to an MPSGD algorithm,an MLF model outperforms the existing state-of-the-art ones in both computational efficiency and scalability.展开更多
To elucidate the geographical differentiation characteristics and driving mechanisms of Dissolved Organic Matter(DOM)in typical rivers,this study conducted a multi-spectral investigation on three representative river ...To elucidate the geographical differentiation characteristics and driving mechanisms of Dissolved Organic Matter(DOM)in typical rivers,this study conducted a multi-spectral investigation on three representative river types within Shandong Province:The mountainous Dawen River,the plain Tuhai River,and the artificial East Grand Canal.The DOM composition was analyzed using Ultraviolet-Visible(UV-Vis)absorption spectroscopy,Excitation-Emission Matrix(EEM)fluorescence spectroscopy,and parallel factor analysis(PARAFAC),while Principal Component Analysis(PCA)was employed to quantify the synergistic effects of natural processes and anthropogenic activities.Results revealed significant spatial heterogeneity in DOM composition and sources.The plain river exhibited the highest aromaticity(humic-like components:43.3%)due to long-term agricultural non-point source inputs and urban wastewater discharge.The mountain stream,shaped by complex terrain and relatively intact ecosystems,was dominated by autochthonous DOM derived from microbial metabolism,with higher Fluorescence Index(FI=2.12)and biological index(BIX=1.35)than other river types.The artificial canal retained protein-like components(64.2%),largely attributed to winter hydrological stagnation and disturbances from shipping activities.Further analysis demonstrated that geographical settings(e.g.,mountain terrain)and anthropogenic activities(e.g.,agriculture,shipping)jointly regulated DOM composition by altering the balance between input and transformation processes.Integrated fluorescence parameters and PCA results suggested differentiated management strategies:protecting ecological integrity in mountain streams to sustain selfpurification,enhancing non-point source interception in plain rivers,and mitigating shipping pollution in canals.This study systematically reveals the natural-anthropogenic coupling mechanisms driving DOM dynamics in northern China rivers,providing critical insights for precision water environment management at the watershed scale.展开更多
A novel algorithm is presented in this paper to jointly estimate frequency,two-dimensional(2-D)direc-tion-of-arrival(DOA),and range of near-field narrowband sources.The proposed method extends the parallel factor(PARA...A novel algorithm is presented in this paper to jointly estimate frequency,two-dimensional(2-D)direc-tion-of-arrival(DOA),and range of near-field narrowband sources.The proposed method extends the parallel factor(PARAFAC)analysis model from the common data-and subspace-domain to the cumu-lant one,and forms three-way arrays by using the five cumulant matrices obtained from the array out-puts,and analyzes the uniqueness of low-rank decomposition of the three-way arrays,then jointly es-timates source parameters via the low-rank decomposition.In comparison with the conventional methods,the proposed method alleviates the loss of the array aperture,and avoids pairing parameters.What is more important,this algorithm can deal with mixed far-field and near-field sources.Finally,the simulation results validate the performance of the proposed method.展开更多
基金supported in part by the National Key Research and Development Program of China(2024YFF0908200)the National Natural Science Foundation of China(62302402,62272078)+1 种基金the Chongqing Natural Science Foundation(CSTB2024TIAD-KPX0018,CSTB2023NSCO-LZX006)the Southwest University Graduate Research Innovation Project(SWUB24050)
摘要High-dimensional and incomplete(HDI) matrices are commonly encountered in various big data-related applications for illustrating the complex interactions among numerous entities, like the user-item interactions in a commercial recommender system or the user-user interactions in a social network services system. The factorization of such an HDI matrix can embed the involved entities into the low-dimensional feature space for acquiring their principal representation, which is a vital task in various application scenes and is often established through the Latent Factor Analysis(LFA). Nevertheless, an HDI matrix can be huge when the corresponding application explodes to involve millions of users, items, or other interactive nodes. In this case, a parallel optimization algorithm is desired for raising the scalability and time efficiency of an LFA model. This paper provides a comprehensive review of the existing parallel optimization algorithms for the LFA model. Specifically, it performs: 1) discussion and summary of these algorithms based on computing architecture and mode, 2) empirical studies of representative models, and3) summary of the current challenges and future directions in this domain. This survey aims to offer an exhaustive review of Parallel Optimization Algorithms for High-Dimensional and Incomplete Matrix Factorization, thereby fostering further research in this field.
摘要A spacecraft attitude estimation method based on electromagnetic vector sensors(EMVS)array is proposed,which employs the orthogonally constrained parallel factor(PARAFAC)algorithm and makes use of measurements of the two-dimensional direction-of-arrival(2D-DOA)and polarization angles,aiming to address the issues of incomplete,asynchronous,and inaccurate third-party reference used for attitude estimation in spacecraft docking missions by employing the electromagnetic wave’s three-dimensional(3D)wave structure as a complete third-party reference.Comparative analysis with state-ofthe-art algorithms shows significant improvements in estimation accuracy and computational efficiency with this algorithm.Numerical simulations have verified the effectiveness and superiority of this method.A high-precision,reliable,and cost-effective method for rapid spacecraft attitude estimation is provided in this paper.
基金supported in part by the National Natural Science Foundation of China(61772493)the Deanship of Scientific Research(DSR)at King Abdulaziz University(RG-48-135-40)+1 种基金Guangdong Province Universities and College Pearl River Scholar Funded Scheme(2019)the Natural Science Foundation of Chongqing(cstc2019jcyjjqX0013)。
摘要A recommender system(RS)relying on latent factor analysis usually adopts stochastic gradient descent(SGD)as its learning algorithm.However,owing to its serial mechanism,an SGD algorithm suffers from low efficiency and scalability when handling large-scale industrial problems.Aiming at addressing this issue,this study proposes a momentum-incorporated parallel stochastic gradient descent(MPSGD)algorithm,whose main idea is two-fold:a)implementing parallelization via a novel datasplitting strategy,and b)accelerating convergence rate by integrating momentum effects into its training process.With it,an MPSGD-based latent factor(MLF)model is achieved,which is capable of performing efficient and high-quality recommendations.Experimental results on four high-dimensional and sparse matrices generated by industrial RS indicate that owing to an MPSGD algorithm,an MLF model outperforms the existing state-of-the-art ones in both computational efficiency and scalability.
基金supported by the National Natural Science Foundation(42472325)the Fundamental Research Funds of Chinese Academy of Geological Science(SK202103).
摘要To elucidate the geographical differentiation characteristics and driving mechanisms of Dissolved Organic Matter(DOM)in typical rivers,this study conducted a multi-spectral investigation on three representative river types within Shandong Province:The mountainous Dawen River,the plain Tuhai River,and the artificial East Grand Canal.The DOM composition was analyzed using Ultraviolet-Visible(UV-Vis)absorption spectroscopy,Excitation-Emission Matrix(EEM)fluorescence spectroscopy,and parallel factor analysis(PARAFAC),while Principal Component Analysis(PCA)was employed to quantify the synergistic effects of natural processes and anthropogenic activities.Results revealed significant spatial heterogeneity in DOM composition and sources.The plain river exhibited the highest aromaticity(humic-like components:43.3%)due to long-term agricultural non-point source inputs and urban wastewater discharge.The mountain stream,shaped by complex terrain and relatively intact ecosystems,was dominated by autochthonous DOM derived from microbial metabolism,with higher Fluorescence Index(FI=2.12)and biological index(BIX=1.35)than other river types.The artificial canal retained protein-like components(64.2%),largely attributed to winter hydrological stagnation and disturbances from shipping activities.Further analysis demonstrated that geographical settings(e.g.,mountain terrain)and anthropogenic activities(e.g.,agriculture,shipping)jointly regulated DOM composition by altering the balance between input and transformation processes.Integrated fluorescence parameters and PCA results suggested differentiated management strategies:protecting ecological integrity in mountain streams to sustain selfpurification,enhancing non-point source interception in plain rivers,and mitigating shipping pollution in canals.This study systematically reveals the natural-anthropogenic coupling mechanisms driving DOM dynamics in northern China rivers,providing critical insights for precision water environment management at the watershed scale.
摘要A novel algorithm is presented in this paper to jointly estimate frequency,two-dimensional(2-D)direc-tion-of-arrival(DOA),and range of near-field narrowband sources.The proposed method extends the parallel factor(PARAFAC)analysis model from the common data-and subspace-domain to the cumu-lant one,and forms three-way arrays by using the five cumulant matrices obtained from the array out-puts,and analyzes the uniqueness of low-rank decomposition of the three-way arrays,then jointly es-timates source parameters via the low-rank decomposition.In comparison with the conventional methods,the proposed method alleviates the loss of the array aperture,and avoids pairing parameters.What is more important,this algorithm can deal with mixed far-field and near-field sources.Finally,the simulation results validate the performance of the proposed method.