Pub Date : 2020-07-06DOI: 10.1108/pap-03-2020-0012
V. Lakshmikantham
{"title":"Preface to special issue","authors":"V. Lakshmikantham","doi":"10.1108/pap-03-2020-0012","DOIUrl":"https://doi.org/10.1108/pap-03-2020-0012","url":null,"abstract":"","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132588875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2012-09-11DOI: 10.1115/1.802977.paper114
M. Soliman, Ghada Y. Abozaid
In cryptography, the advanced encryption standard (AES) is an encryption standard issued as FIPS by NIST as a successor to data encryption standard (DES) algorithm. The applications of the AES are wide including any sensitive data that requires cryptographic protection before communication or storage. This paper proposes extending general-purpose processors with crypto coprocessor based on decoupled architectures. The extended coprocessor splits an encryption/decryption instruction into memory (load/store) and computation (encryption/decryption) portions (pseudo instructions). Loading/storing and encrypting/decrypting data are performed in parallel and communicated through architectural queues. The computational unit includes parallel AES pipelines for fast encrypting/decrypting data. On four parallel AES pipelines, our results show a performance of 222 Giga bits per second.
{"title":"FastCrypto: parallel AES pipelines extension for general-purpose processors","authors":"M. Soliman, Ghada Y. Abozaid","doi":"10.1115/1.802977.paper114","DOIUrl":"https://doi.org/10.1115/1.802977.paper114","url":null,"abstract":"In cryptography, the advanced encryption standard (AES) is an encryption standard issued as FIPS by NIST as a successor to data encryption standard (DES) algorithm. The applications of the AES are wide including any sensitive data that requires cryptographic protection before communication or storage. This paper proposes extending general-purpose processors with crypto coprocessor based on decoupled architectures. The extended coprocessor splits an encryption/decryption instruction into memory (load/store) and computation (encryption/decryption) portions (pseudo instructions). Loading/storing and encrypting/decrypting data are performed in parallel and communicated through architectural queues. The computational unit includes parallel AES pipelines for fast encrypting/decrypting data. On four parallel AES pipelines, our results show a performance of 222 Giga bits per second.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2012-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122563332","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The recent high rate of globalization of new disease strains and infectious agents at non-endemic zones is closely associated with complex human population structure and the large-scale inter-patch connections human transportations. The complexities in the human population structure create heterogeneities with respect to patch dwelling populations as well as endemic population structure. We present a multi-group SIRS dynamic epidemic process in the context of scale structured population. For simplicity we limit the scale to the value of two and define a multi-scale extension algorithm for the epidemic process. We investigate the global uniform asymptotic stability of the disease free equilibrium of the scale structured epidemic dynamic process and its impact on the emergence, propagation and resurgence of the disease. The presented results are demonstrated by numerical simulation results.
{"title":"Global stability of two-scale network human epidemic dynamic model","authors":"D. Wanduku, G. Ladde","doi":"10.5555/2336431.2336436","DOIUrl":"https://doi.org/10.5555/2336431.2336436","url":null,"abstract":"The recent high rate of globalization of new disease strains and infectious agents at non-endemic zones is closely associated with complex human population structure and the large-scale inter-patch connections human transportations. The complexities in the human population structure create heterogeneities with respect to patch dwelling populations as well as endemic population structure. We present a multi-group SIRS dynamic epidemic process in the context of scale structured population. For simplicity we limit the scale to the value of two and define a multi-scale extension algorithm for the epidemic process. We investigate the global uniform asymptotic stability of the disease free equilibrium of the scale structured epidemic dynamic process and its impact on the emergence, propagation and resurgence of the disease. The presented results are demonstrated by numerical simulation results.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2011-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114635265","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The aim of this paper is to numerically investigate a PH/PH/1 inventory model with reneging of customers and finite shortage of items. We assume that arrivals occur according to a phase type renewal process. The associated phase type distribution has representation (α, U). The service times are identically and independently distributed random variables having common phase type distribution with representation (β, V). The lead-time is zero. Costumers renege from the system at a constant rate γ. Shortage is permitted and hence shortage cost is finite. We perform the steady state analysis of the inventory model using Matrix analytic method. A suitable cost function is defined and analyzed numerically. The optimal shortage level is numerically evaluated. Some measures of the system performance in the steady state are also derived.
{"title":"Numerical investigation of a PH/PH/1 inventory system with positive service time and shortage","authors":"A. Krishnamoorthy, K. P. Jose, V. C. Narayanan","doi":"10.5555/1561721.1561730","DOIUrl":"https://doi.org/10.5555/1561721.1561730","url":null,"abstract":"The aim of this paper is to numerically investigate a PH/PH/1 inventory model with reneging of customers and finite shortage of items. We assume that arrivals occur according to a phase type renewal process. The associated phase type distribution has representation (α, U). The service times are identically and independently distributed random variables having common phase type distribution with representation (β, V). The lead-time is zero. Costumers renege from the system at a constant rate γ. Shortage is permitted and hence shortage cost is finite. We perform the steady state analysis of the inventory model using Matrix analytic method. A suitable cost function is defined and analyzed numerically. The optimal shortage level is numerically evaluated. Some measures of the system performance in the steady state are also derived.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2008-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133571135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Back Propagation (BP) training algorithm has received intensive research efforts to exploit its parallelism in order to reduce the training time for complex problems. A modified version of BP based on matrix-matrix multiplication was proposed for parallel processing. This paper discusses the implementation of Matrix Back Propagation (MBP) using scalar, vector, and matrix instruction set architecture (ISA). Besides, it shows that the performance of the MBP is improved by switching form scalar to vector ISA and form vector to matrix ISA. On a practical application, speech recognition, the speedup of training a neural network using unrolling scalar over scalar ISA is 1.83. On eight parallel lanes, the speedup of using vector, unrolling vector, and matrix ISA are respectively 10.33, 11.88, and 15.36, where the maximum theoretical speedup is 16. Our results show that the use of matrix ISA gives a performance close to the optimal because of reusing the loaded data, decreasing the loop overhead, and overlapping the memory operations by arithmetic operations.
{"title":"A highly efficient implementation of back propagation algorithm using matrix instruction set architecture","authors":"M. Soliman, S. Mohamed","doi":"10.5555/1315424.1315425","DOIUrl":"https://doi.org/10.5555/1315424.1315425","url":null,"abstract":"Back Propagation (BP) training algorithm has received intensive research efforts to exploit its parallelism in order to reduce the training time for complex problems. A modified version of BP based on matrix-matrix multiplication was proposed for parallel processing. This paper discusses the implementation of Matrix Back Propagation (MBP) using scalar, vector, and matrix instruction set architecture (ISA). Besides, it shows that the performance of the MBP is improved by switching form scalar to vector ISA and form vector to matrix ISA. On a practical application, speech recognition, the speedup of training a neural network using unrolling scalar over scalar ISA is 1.83. On eight parallel lanes, the speedup of using vector, unrolling vector, and matrix ISA are respectively 10.33, 11.88, and 15.36, where the maximum theoretical speedup is 16. Our results show that the use of matrix ISA gives a performance close to the optimal because of reusing the loaded data, decreasing the loop overhead, and overlapping the memory operations by arithmetic operations.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2007-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121007161","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Events generated by the execution of a distributed system are related by causality and concurrency. While providing a means of reasoning about the relative occurrence of events, this partial order fails to represent the timeliness of occurrence. In this paper, we develop a novel means of assigning weights to events where the weights are reduced as the temporal proximity to an anchor event decreases. This weight quantifies the strength of the causal or concurrent relationship with respect to an anchor event. Those events that causally succeed the anchor are the focus of this paper with concurrency and causally preceding being part of future work plans. Three methods of computing event weights for causally succeeding events are defined. Each contains a tunable parameter to determine the rate of weight decrease. The methods are piece-wise linear, exponential, and relevant vector difference decay. A case study has been performed that applied quantitative causality to the well-known software engineering problem of feature location. A summary of the case study results is provided to illustrate the utility of quantitative causality for succeeding events.
{"title":"Quantitative causality","authors":"S. Simmons, Dennis Edwards","doi":"10.1063/pt.5.7124","DOIUrl":"https://doi.org/10.1063/pt.5.7124","url":null,"abstract":"Events generated by the execution of a distributed system are related by causality and concurrency. While providing a means of reasoning about the relative occurrence of events, this partial order fails to represent the timeliness of occurrence. In this paper, we develop a novel means of assigning weights to events where the weights are reduced as the temporal proximity to an anchor event decreases. This weight quantifies the strength of the causal or concurrent relationship with respect to an anchor event. Those events that causally succeed the anchor are the focus of this paper with concurrency and causally preceding being part of future work plans. Three methods of computing event weights for causally succeeding events are defined. Each contains a tunable parameter to determine the rate of weight decrease. The methods are piece-wise linear, exponential, and relevant vector difference decay. A case study has been performed that applied quantitative causality to the well-known software engineering problem of feature location. A summary of the case study results is provided to illustrate the utility of quantitative causality for succeeding events.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2007-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127259936","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2001-06-01DOI: 10.1142/9789812778086_0012
A. Kastania, M. P. Bekakos
Speed of execution is an important issue in applying an Artificial Neural Network (ANN) to any real-time problem. Adaptive Logic Networks (ALNs) are a type of ANNs, which allow the solution of pattern classification problems at very high speed based on traversal of a binary decision tree architecture. Herein, various ALN architectures were built and evaluated based on the relational approach for ALNs. The outcome was the development of an integrated intelligent classification engine (I2CE) for biosignal engineering purposes.
{"title":"Integrated intelligent classification engine (I2CE) for biosignal engineering","authors":"A. Kastania, M. P. Bekakos","doi":"10.1142/9789812778086_0012","DOIUrl":"https://doi.org/10.1142/9789812778086_0012","url":null,"abstract":"Speed of execution is an important issue in applying an Artificial Neural Network (ANN) to any real-time problem. Adaptive Logic Networks (ALNs) are a type of ANNs, which allow the solution of pattern classification problems at very high speed based on traversal of a binary decision tree architecture. Herein, various ALN architectures were built and evaluated based on the relational approach for ALNs. The outcome was the development of an integrated intelligent classification engine (I2CE) for biosignal engineering purposes.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2001-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128575212","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1996-03-01DOI: 10.1080/10637199608915573
Qiming He, D. J. Evans, Lishan Kang
In this paper, a new monotone-Schwarz parallel algorithm for solving a class of semilinear elliptic systems is proposed. In the case of overlapping subdomains, the detailed procedures for constructing iterative sequences and the convergence proofs are investigated.
{"title":"Monotone-Schwarz parallel algorithm for nonlinear elliptic equations","authors":"Qiming He, D. J. Evans, Lishan Kang","doi":"10.1080/10637199608915573","DOIUrl":"https://doi.org/10.1080/10637199608915573","url":null,"abstract":"In this paper, a new monotone-Schwarz parallel algorithm for solving a class of semilinear elliptic systems is proposed. In the case of overlapping subdomains, the detailed procedures for constructing iterative sequences and the convergence proofs are investigated.","PeriodicalId":212567,"journal":{"name":"Neural Parallel Sci. Comput.","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"1996-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124689441","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：