Deep matrix factorization(DMF)has been demonstrated to be a powerful tool to take in the complex hierarchical information of multi-view data(MDR).However,existing multiview DMF methods mainly explore the consistency o...Deep matrix factorization(DMF)has been demonstrated to be a powerful tool to take in the complex hierarchical information of multi-view data(MDR).However,existing multiview DMF methods mainly explore the consistency of multi-view data,while neglecting the diversity among different views as well as the high-order relationships of data,resulting in the loss of valuable complementary information.In this paper,we design a hypergraph regularized diverse deep matrix factorization(HDDMF)model for multi-view data representation,to jointly utilize multi-view diversity and a high-order manifold in a multilayer factorization framework.A novel diversity enhancement term is designed to exploit the structural complementarity between different views of data.Hypergraph regularization is utilized to preserve the high-order geometry structure of data in each view.An efficient iterative optimization algorithm is developed to solve the proposed model with theoretical convergence analysis.Experimental results on five real-world data sets demonstrate that the proposed method significantly outperforms stateof-the-art multi-view learning approaches.展开更多
The parametric temporal data model captures a real world entity in a single tuple, which reduces query language complexity. Such a data model, however, is difficult to be implemented on top of conventional databases b...The parametric temporal data model captures a real world entity in a single tuple, which reduces query language complexity. Such a data model, however, is difficult to be implemented on top of conventional databases because of its unfixed attribute sizes. XML is a matured technology and can be an elegant solution for such challenge. Representing data in XML trigger a question about storage efficiency. The goal of this work is to provide a straightforward answer to such a question. To this end, we compare three different storage models for the parametric temporal data model and show that XML is not worse than any other approaches. Furthermore, XML outperforms the other storages under certain conditions. Therefore, our simulation results provide a positive indication that the myth about XML is not true in the parametric temporal data model.展开更多
Society is increasingly relying on artificially intelligent(AI) systems to facilitate, and sometimes even automate, critical systems that have huge impacts on the people these systems are designed to serve. But the un...Society is increasingly relying on artificially intelligent(AI) systems to facilitate, and sometimes even automate, critical systems that have huge impacts on the people these systems are designed to serve. But the unique nature of AI systems opens up new challenges regarding their ethical use. For example,(1) unrepresentative training data can introduce sampling bias, leading to unfair outcomes;and(2) lack of data equity can introduce systemic bias into the system. At the university level, how to provide ethics training within the limits of typical computer science(CS) programs is non-trivial, as current CS education programs already face heavy burdens from unprecedented demand. In particular, the University of Michigan is exploring ways of introducing ethics training within the CS curricula, including both stand-alone courses and integrated modules.展开更多
Big data analytic techniques associated with machine learning algorithms are playing an increasingly important role in various application fields,including stock market investment.However,few studies have focused on f...Big data analytic techniques associated with machine learning algorithms are playing an increasingly important role in various application fields,including stock market investment.However,few studies have focused on forecasting daily stock market returns,especially when using powerful machine learning techniques,such as deep neural networks(DNNs),to perform the analyses.DNNs employ various deep learning algorithms based on the combination of network structure,activation function,and model parameters,with their performance depending on the format of the data representation.This paper presents a comprehensive big data analytics process to predict the daily return direction of the SPDR S&P 500 ETF(ticker symbol:SPY)based on 60 financial and economic features.DNNs and traditional artificial neural networks(ANNs)are then deployed over the entire preprocessed but untransformed dataset,along with two datasets transformed via principal component analysis(PCA),to predict the daily direction of future stock market index returns.While controlling for overfitting,a pattern for the classification accuracy of the DNNs is detected and demonstrated as the number of the hidden layers increases gradually from 12 to 1000.Moreover,a set of hypothesis testing procedures are implemented on the classification,and the simulation results show that the DNNs using two PCA-represented datasets give significantly higher classification accuracy than those using the entire untransformed dataset,as well as several other hybrid machine learning algorithms.In addition,the trading strategies guided by the DNN classification process based on PCA-represented data perform slightly better than the others tested,including in a comparison against two standard benchmarks.展开更多
Deep learning algorithms increasingly support automated systems in areas such as human activity recognition and purchase recommendation.We identify a current trend in which data is transformed first into abstract visu...Deep learning algorithms increasingly support automated systems in areas such as human activity recognition and purchase recommendation.We identify a current trend in which data is transformed first into abstract visualizations and then processed by a computer vision deep learning pipeline.We call this VisuaLization As Intermediate Representation(VLAIR)and believe that it can be instrumental to support accurate recognition in a number of fields while also enhancing humans’ability to interpret deep learning models for debugging purposes or for personal use.In this paper we describe the potential advantages of this approach and explore various visualization mappings and deep learning architectures.We evaluate several VLAIR alternatives for a specific problem(human activity recognition in an apartment)and show that VLAIR attains classification accuracy above classical machine learning algorithms and several other non-image-based deep learning algorithms with several data representations.展开更多
Modern underwater object detection methods recognize objects from sonar data based on their geometric shapes.However,the distortion of objects during data acquisition and representation is seldom considered.In this pa...Modern underwater object detection methods recognize objects from sonar data based on their geometric shapes.However,the distortion of objects during data acquisition and representation is seldom considered.In this paper,we present a detailed summary of representations for sonar data and a concrete analysis of the geometric characteristics of different data representations.Based on this,a feature fusion framework is proposed to fully use the intensity features extracted from the polar image representation and the geometric features learned from the point cloud representation of sonar data.Three feature fusion strategies are presented to investigate the impact of feature fusion on different components of the detection pipeline.In addition,the fusion strategies can be easily integrated into other detectors,such as the You Only Look Once(YOLO)series.The effectiveness of our proposed framework and feature fusion strategies is demonstrated on a public sonar dataset captured in real-world underwater environments.Experimental results show that our method benefits both the region proposal and the object classification modules in the detectors.展开更多
In response to climate change,there has been a focus on developing lightweight and environmentally friendly materials,with active research aimed at enhancing the energy efficiency of electric and hybrid vehicles.In th...In response to climate change,there has been a focus on developing lightweight and environmentally friendly materials,with active research aimed at enhancing the energy efficiency of electric and hybrid vehicles.In this context,the development of polymer composites with superior thermal conductivity(TC)has been recognized as critical to meeting mechanical property requirements.This paper presents a machine learning model that utilized 1774 experimental data points to predict various properties of polymer composites,such as density,heat deflection temperature,flexural modulus,flexural strength,tensile yield strength,impact strength,and TC.Various data representation methods for composition data are employed,and the XGBoost model is trained,achieving high accuracy with an average R2 score of 0.95.This machine learning model,informed by experimental data,is a useful tool for predicting and optimizing the properties of polymer composites.展开更多
Powerful expressive ability of semantic information, to be easily computed and flexibility are basic features of digital product model (DPM). Using ontology and object-oriented principle (OOP) together to cope with pr...Powerful expressive ability of semantic information, to be easily computed and flexibility are basic features of digital product model (DPM). Using ontology and object-oriented principle (OOP) together to cope with problems in modeling is brought forward in this paper. The two are widely used and do well in modeling, but they each alone cannot cope with all issues and new challenges. Three basic requests are pointed out in DPM modeling. Status, problems, and root of current non-semantic and semantic models are introduced. Ontology, OOP, and their difference are introduced. It is found that the two are entirely complementary with each other. How to assign the roles and to cooperate for the two in coping with the three basic issues in DPM modeling are explained in detail.展开更多
As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing wi...As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing with high hardware efficiency has been considered for implementing the computation units of CNN accelerators.However,individual approximate designs such as multipliers and adders can only achieve limited accuracy and hardware improvements.In this paper,an approximate PE is dedicatedly devised for CNN accelerators by synergistically considering the data representation,multiplication and accumulation.An approximate data format is defined for the weights using stochastic rounding.This data format enables a simple implementation of multiplication by using small lookup tables,an adder and a shifter.Two approximate accumulators are further proposed for the product accumulation in the PE.Compared with the exact 8-bit fixed-point design,the proposed PE saves more than 29%and 20%in power-delay product for 3×3 and 5×5 sum of products,respectively.Also,compared with the PEs consisting of state-of-the-art approximate multipliers,the proposed design shows significantly smaller error bias with lower hardware overhead.Moreover,the application of the approximate PEs in CNN accelerators is analyzed by implementing a multi-task CNN for face detection and alignment.We conclude that 1)an approximate PE is more effective for face detection than for alignment,2)an approximate PE with high statistically-measured accuracy does not necessarily result in good quality in face detection,and 3)properly increasing the number of PEs in a CNN accelerator can improve its power and energy efficiency.展开更多
Vast amounts of data are produced with the development of smart cities and urban computing technologies.The data is often captured from multiple sensors,with heterogeneous structures and highly decentralized connectio...Vast amounts of data are produced with the development of smart cities and urban computing technologies.The data is often captured from multiple sensors,with heterogeneous structures and highly decentralized connections.Integrated data representation and smart computational models are required for more complex tasks in urban computing.We dwell deeply on two fundamental questions—can we provide an integrated data representation for the whole cyber–physical–social system?And,can we provide an integrated framework to choose the appropriate data for understanding a specific urban event?A holography data representation and the quasi-holography computational model have been proposed to address these problems.We describe case studies using the quasi-holography computational model,and discuss further problems to solve regarding our model.展开更多
In real-world driving scenarios,understanding the intentions of pedestrians in real-time is critical for the built environment safety when operating intelligent vehicles on the roads.Pedestrians crossing the street is...In real-world driving scenarios,understanding the intentions of pedestrians in real-time is critical for the built environment safety when operating intelligent vehicles on the roads.Pedestrians crossing the street is a common behavior that can easily lead to accidents.This paper presents a comprehensive review of the prediction of pedestrian crossing intentions,focusing on data,model structure,data representation,information extraction,prediction function,and associated models and challenges.The review highlights that data types,model generalization ability,and prediction uncertainty are key challenges on pedestrian crossing intention prediction.It identifies open challenges and opportunities for future research in pedestrian crossing intention prediction.展开更多
基金This work was supported by the National Natural Science Foundation of China(62073087,62071132,61973090).
文摘Deep matrix factorization(DMF)has been demonstrated to be a powerful tool to take in the complex hierarchical information of multi-view data(MDR).However,existing multiview DMF methods mainly explore the consistency of multi-view data,while neglecting the diversity among different views as well as the high-order relationships of data,resulting in the loss of valuable complementary information.In this paper,we design a hypergraph regularized diverse deep matrix factorization(HDDMF)model for multi-view data representation,to jointly utilize multi-view diversity and a high-order manifold in a multilayer factorization framework.A novel diversity enhancement term is designed to exploit the structural complementarity between different views of data.Hypergraph regularization is utilized to preserve the high-order geometry structure of data in each view.An efficient iterative optimization algorithm is developed to solve the proposed model with theoretical convergence analysis.Experimental results on five real-world data sets demonstrate that the proposed method significantly outperforms stateof-the-art multi-view learning approaches.
基金supported by the National Research Foundation in Korea through contract N-12-NM-IR05
文摘The parametric temporal data model captures a real world entity in a single tuple, which reduces query language complexity. Such a data model, however, is difficult to be implemented on top of conventional databases because of its unfixed attribute sizes. XML is a matured technology and can be an elegant solution for such challenge. Representing data in XML trigger a question about storage efficiency. The goal of this work is to provide a straightforward answer to such a question. To this end, we compare three different storage models for the parametric temporal data model and show that XML is not worse than any other approaches. Furthermore, XML outperforms the other storages under certain conditions. Therefore, our simulation results provide a positive indication that the myth about XML is not true in the parametric temporal data model.
文摘Society is increasingly relying on artificially intelligent(AI) systems to facilitate, and sometimes even automate, critical systems that have huge impacts on the people these systems are designed to serve. But the unique nature of AI systems opens up new challenges regarding their ethical use. For example,(1) unrepresentative training data can introduce sampling bias, leading to unfair outcomes;and(2) lack of data equity can introduce systemic bias into the system. At the university level, how to provide ethics training within the limits of typical computer science(CS) programs is non-trivial, as current CS education programs already face heavy burdens from unprecedented demand. In particular, the University of Michigan is exploring ways of introducing ethics training within the CS curricula, including both stand-alone courses and integrated modules.
文摘Big data analytic techniques associated with machine learning algorithms are playing an increasingly important role in various application fields,including stock market investment.However,few studies have focused on forecasting daily stock market returns,especially when using powerful machine learning techniques,such as deep neural networks(DNNs),to perform the analyses.DNNs employ various deep learning algorithms based on the combination of network structure,activation function,and model parameters,with their performance depending on the format of the data representation.This paper presents a comprehensive big data analytics process to predict the daily return direction of the SPDR S&P 500 ETF(ticker symbol:SPY)based on 60 financial and economic features.DNNs and traditional artificial neural networks(ANNs)are then deployed over the entire preprocessed but untransformed dataset,along with two datasets transformed via principal component analysis(PCA),to predict the daily direction of future stock market index returns.While controlling for overfitting,a pattern for the classification accuracy of the DNNs is detected and demonstrated as the number of the hidden layers increases gradually from 12 to 1000.Moreover,a set of hypothesis testing procedures are implemented on the classification,and the simulation results show that the DNNs using two PCA-represented datasets give significantly higher classification accuracy than those using the entire untransformed dataset,as well as several other hybrid machine learning algorithms.In addition,the trading strategies guided by the DNN classification process based on PCA-represented data perform slightly better than the others tested,including in a comparison against two standard benchmarks.
文摘Deep learning algorithms increasingly support automated systems in areas such as human activity recognition and purchase recommendation.We identify a current trend in which data is transformed first into abstract visualizations and then processed by a computer vision deep learning pipeline.We call this VisuaLization As Intermediate Representation(VLAIR)and believe that it can be instrumental to support accurate recognition in a number of fields while also enhancing humans’ability to interpret deep learning models for debugging purposes or for personal use.In this paper we describe the potential advantages of this approach and explore various visualization mappings and deep learning architectures.We evaluate several VLAIR alternatives for a specific problem(human activity recognition in an apartment)and show that VLAIR attains classification accuracy above classical machine learning algorithms and several other non-image-based deep learning algorithms with several data representations.
基金supported by the National Natural Science Foundation of China(No.62103072)the Postdoctoral Science Foundation of China(No.2021M690502)。
文摘Modern underwater object detection methods recognize objects from sonar data based on their geometric shapes.However,the distortion of objects during data acquisition and representation is seldom considered.In this paper,we present a detailed summary of representations for sonar data and a concrete analysis of the geometric characteristics of different data representations.Based on this,a feature fusion framework is proposed to fully use the intensity features extracted from the polar image representation and the geometric features learned from the point cloud representation of sonar data.Three feature fusion strategies are presented to investigate the impact of feature fusion on different components of the detection pipeline.In addition,the fusion strategies can be easily integrated into other detectors,such as the You Only Look Once(YOLO)series.The effectiveness of our proposed framework and feature fusion strategies is demonstrated on a public sonar dataset captured in real-world underwater environments.Experimental results show that our method benefits both the region proposal and the object classification modules in the detectors.
基金supported by the Ministry of Trade,Industry and Energy(MOTIE),Korea Institute for Advancement of Technology(KIAT)through the Virtual Engineering Platform Program(P0022334)Korea Planning and Evaluation Institute of Industrial Technology(KEIT)R&D Program(No.2000391).
文摘In response to climate change,there has been a focus on developing lightweight and environmentally friendly materials,with active research aimed at enhancing the energy efficiency of electric and hybrid vehicles.In this context,the development of polymer composites with superior thermal conductivity(TC)has been recognized as critical to meeting mechanical property requirements.This paper presents a machine learning model that utilized 1774 experimental data points to predict various properties of polymer composites,such as density,heat deflection temperature,flexural modulus,flexural strength,tensile yield strength,impact strength,and TC.Various data representation methods for composition data are employed,and the XGBoost model is trained,achieving high accuracy with an average R2 score of 0.95.This machine learning model,informed by experimental data,is a useful tool for predicting and optimizing the properties of polymer composites.
基金Supported by the Ministries’ Basic Research Foundation, China Knowledge Engineering Platform for Enterprise Innovative Design(No. B0920060901)
文摘Powerful expressive ability of semantic information, to be easily computed and flexibility are basic features of digital product model (DPM). Using ontology and object-oriented principle (OOP) together to cope with problems in modeling is brought forward in this paper. The two are widely used and do well in modeling, but they each alone cannot cope with all issues and new challenges. Three basic requests are pointed out in DPM modeling. Status, problems, and root of current non-semantic and semantic models are introduced. Ontology, OOP, and their difference are introduced. It is found that the two are entirely complementary with each other. How to assign the roles and to cooperate for the two in coping with the three basic issues in DPM modeling are explained in detail.
基金supported in part by the National Natural Science Foundation of China under Grant No.62104127the National Key Research and Development Program of China under Grant No.2022YFB4500200.
文摘As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing with high hardware efficiency has been considered for implementing the computation units of CNN accelerators.However,individual approximate designs such as multipliers and adders can only achieve limited accuracy and hardware improvements.In this paper,an approximate PE is dedicatedly devised for CNN accelerators by synergistically considering the data representation,multiplication and accumulation.An approximate data format is defined for the weights using stochastic rounding.This data format enables a simple implementation of multiplication by using small lookup tables,an adder and a shifter.Two approximate accumulators are further proposed for the product accumulation in the PE.Compared with the exact 8-bit fixed-point design,the proposed PE saves more than 29%and 20%in power-delay product for 3×3 and 5×5 sum of products,respectively.Also,compared with the PEs consisting of state-of-the-art approximate multipliers,the proposed design shows significantly smaller error bias with lower hardware overhead.Moreover,the application of the approximate PEs in CNN accelerators is analyzed by implementing a multi-task CNN for face detection and alignment.We conclude that 1)an approximate PE is more effective for face detection than for alignment,2)an approximate PE with high statistically-measured accuracy does not necessarily result in good quality in face detection,and 3)properly increasing the number of PEs in a CNN accelerator can improve its power and energy efficiency.
基金This work was supported by National Basic Research Program of China(973 Program)(2015CB352501,2015CB352500).
文摘Vast amounts of data are produced with the development of smart cities and urban computing technologies.The data is often captured from multiple sensors,with heterogeneous structures and highly decentralized connections.Integrated data representation and smart computational models are required for more complex tasks in urban computing.We dwell deeply on two fundamental questions—can we provide an integrated data representation for the whole cyber–physical–social system?And,can we provide an integrated framework to choose the appropriate data for understanding a specific urban event?A holography data representation and the quasi-holography computational model have been proposed to address these problems.We describe case studies using the quasi-holography computational model,and discuss further problems to solve regarding our model.
文摘In real-world driving scenarios,understanding the intentions of pedestrians in real-time is critical for the built environment safety when operating intelligent vehicles on the roads.Pedestrians crossing the street is a common behavior that can easily lead to accidents.This paper presents a comprehensive review of the prediction of pedestrian crossing intentions,focusing on data,model structure,data representation,information extraction,prediction function,and associated models and challenges.The review highlights that data types,model generalization ability,and prediction uncertainty are key challenges on pedestrian crossing intention prediction.It identifies open challenges and opportunities for future research in pedestrian crossing intention prediction.
