The Experts below are selected from a list of 25626 Experts worldwide ranked by ideXlab platform

Salman A Avestimehr - One of the best experts on this subject based on the ideXlab platform.

  • polynomial codes an optimal design for high dimensional coded matrix multiplication
    Neural Information Processing Systems, 2017
    Co-Authors: Mohammad Ali Maddahali, Salman A Avestimehr
    Abstract:

    We consider a large-scale matrix multiplication problem where the computation is carried out using a distributed system with a master node and multiple worker nodes, where each worker can store parts of the input matrices. We propose a computation strategy that leverages ideas from coding theory to design intermediate computations at the worker nodes, in order to optimally deal with straggling workers. The proposed strategy, named as polynomial codes, achieves the Optimum Recovery threshold, defined as the minimum number of workers that the master needs to wait for in order to compute the output. This is the first code that achieves the optimal utilization of redundancy for tolerating stragglers or failures in distributed matrix multiplication. Furthermore, by leveraging the algebraic structure of polynomial codes, we can map the reconstruction problem of the final output to a polynomial interpolation problem, which can be solved efficiently. Polynomial codes provide order-wise improvement over the state of the art in terms of Recovery threshold, and are also optimal in terms of several other metrics including computation latency and communication load. Moreover, we extend this code to distributed convolution and show its order-wise optimality.

  • polynomial codes an optimal design for high dimensional coded matrix multiplication
    arXiv: Information Theory, 2017
    Co-Authors: Mohammad Ali Maddahali, Salman A Avestimehr
    Abstract:

    We consider a large-scale matrix multiplication problem where the computation is carried out using a distributed system with a master node and multiple worker nodes, where each worker can store parts of the input matrices. We propose a computation strategy that leverages ideas from coding theory to design intermediate computations at the worker nodes, in order to efficiently deal with straggling workers. The proposed strategy, named as \emph{polynomial codes}, achieves the Optimum Recovery threshold, defined as the minimum number of workers that the master needs to wait for in order to compute the output. Furthermore, by leveraging the algebraic structure of polynomial codes, we can map the reconstruction problem of the final output to a polynomial interpolation problem, which can be solved efficiently. Polynomial codes provide order-wise improvement over the state of the art in terms of Recovery threshold, and are also optimal in terms of several other metrics. Furthermore, we extend this code to distributed convolution and show its order-wise optimality.

Mohammad Ali Maddahali - One of the best experts on this subject based on the ideXlab platform.

  • polynomial codes an optimal design for high dimensional coded matrix multiplication
    Neural Information Processing Systems, 2017
    Co-Authors: Mohammad Ali Maddahali, Salman A Avestimehr
    Abstract:

    We consider a large-scale matrix multiplication problem where the computation is carried out using a distributed system with a master node and multiple worker nodes, where each worker can store parts of the input matrices. We propose a computation strategy that leverages ideas from coding theory to design intermediate computations at the worker nodes, in order to optimally deal with straggling workers. The proposed strategy, named as polynomial codes, achieves the Optimum Recovery threshold, defined as the minimum number of workers that the master needs to wait for in order to compute the output. This is the first code that achieves the optimal utilization of redundancy for tolerating stragglers or failures in distributed matrix multiplication. Furthermore, by leveraging the algebraic structure of polynomial codes, we can map the reconstruction problem of the final output to a polynomial interpolation problem, which can be solved efficiently. Polynomial codes provide order-wise improvement over the state of the art in terms of Recovery threshold, and are also optimal in terms of several other metrics including computation latency and communication load. Moreover, we extend this code to distributed convolution and show its order-wise optimality.

  • polynomial codes an optimal design for high dimensional coded matrix multiplication
    arXiv: Information Theory, 2017
    Co-Authors: Mohammad Ali Maddahali, Salman A Avestimehr
    Abstract:

    We consider a large-scale matrix multiplication problem where the computation is carried out using a distributed system with a master node and multiple worker nodes, where each worker can store parts of the input matrices. We propose a computation strategy that leverages ideas from coding theory to design intermediate computations at the worker nodes, in order to efficiently deal with straggling workers. The proposed strategy, named as \emph{polynomial codes}, achieves the Optimum Recovery threshold, defined as the minimum number of workers that the master needs to wait for in order to compute the output. Furthermore, by leveraging the algebraic structure of polynomial codes, we can map the reconstruction problem of the final output to a polynomial interpolation problem, which can be solved efficiently. Polynomial codes provide order-wise improvement over the state of the art in terms of Recovery threshold, and are also optimal in terms of several other metrics. Furthermore, we extend this code to distributed convolution and show its order-wise optimality.

S M Ghoreishi - One of the best experts on this subject based on the ideXlab platform.

  • response surface optimization of safranal and crocin extraction from crocus sativus l via supercritical fluid technology
    Journal of Supercritical Fluids, 2016
    Co-Authors: Ghorbandoust M Goleroudbary, S M Ghoreishi
    Abstract:

    Abstract Safranal (2,3-Dihydro-2,2,6-trimethylbenzaldehyde) and Crocin (Crocetin digentoboisyl ester) were experimentally extracted from saffron ( Crocus sativus L. ) via supercritical carbon dioxide (SC-CO 2 ) and modified SC-CO 2 with methanol as entrainer, respectively. The extracted samples were analyzed by high performance liquid chromatography (HPLC). The response surface methodology (RSM) was applied to evaluate the effects of four independent variables (temperature, pressure, SC-CO 2 flow rate, and dynamic extraction time) on the Recovery of Safranal and Crocin. The Recovery results indicated that the data adequately fitted into a second-order polynomial model. The optimal values of variables for Safranal extraction were determined by RSM to be 92 °C, 21.3 MPa, 0.9 cm 3  min −1 and 122.0 min for the Optimum predicted Recovery of 91.76 w/w%. Optimum Recovery of Crocin (32.67 w/w%) was obtained at 44 °C, 19.3 MPa, 1.0 cm 3  min −1 and 110.0 min. The accuracy of the modeling optimal Safranal and Crocin Recovery was validated with triplicate experiments giving the average extraction Recovery of 90.03 and 33.05 respectively.

  • extraction of epigallocatechin 3 gallate from green tea via supercritical fluid technology neural network modeling and response surface optimization
    Journal of Supercritical Fluids, 2013
    Co-Authors: S M Ghoreishi, Elham Heidari
    Abstract:

    Abstract In this study the extraction of (−)-Epigallocatechin-3-gallate (EGCG) from Iranian green tea was investigated by supercritical CO 2 with ethanol as co-solvent. Design of experiments and modeling were carried out with response surface methodology by Minitab software. The HPLC analysis of the extracted samples was used in conjunction with response surface design to optimize four operating variables of supercritical CO 2 extraction (pressure, temperature, CO 2 flow rate and extraction dynamic time). Optimum Recovery of EGCG (0.462 g/g) was obtained at 19.3 MPa, 43.7 °C, 106 min (dynamic) and 1.5 ml/min (CO 2 flow rate). Moreover, a three-layer artificial neural network was developed for modeling EGCG extraction from green tea. In this regard, different networks (by changing the number of neurons in the hidden layer and algorithm of network training) were compared with evaluation of networks accuracy in extraction Recovery prediction. Finally, the Levenberg–Marquardt algorithm with the six neurons in the hidden layer has been found to be the most suitable network.

Gilbert Pepin - One of the best experts on this subject based on the ideXlab platform.

  • compared interest between hair analysis and urinalysis in doping controls results for amphetamines corticosteroids and anabolic steroids in racing cyclists
    Forensic Science International, 2000
    Co-Authors: Yvan Gaillard, Francois Vayssette, Gilbert Pepin
    Abstract:

    Abstract In France during a famous bicycle race, the newspapers documented the degree in which doping seemed to be supervised in some teams by managers and doctors. Use of anabolic steroids and other substances was officially banned in the mid-seventies by sports authorities. This policy has been enforced through urine testing before competition. It is well known, however, that a latency period is all that is necessary to defeat these tests. Nevertheless, hair analysis could be a promising tool when testing for periods that are not accessible to urinalysis any more. We have developed different sensitive methods for testing hair for amphetamines, anabolic steroids and their esters and corticosteroids. For amphetamines, 50 mg of hair were digested with 1 M NaOH, extracted with ethyl acetate, derivatized with TFA and analyzed by gas chromatography positive chemical-ionization mass spectrometry. For corticosteroids, 50 mg of powdered hair were treated with methanol in an ultrasonic bath and subsequently purified using a C 18 solid phase extraction column. Analysis was realized by high performance liquid chromatography coupled to electrospray-ionization tandem mass spectrometry. For anabolic steroids and their esters, 100 mg of powdered hair were treated with methanol in an ultrasonic bath for extraction of esters, then alkaline digested with 1 M NaOH for an Optimum Recovery of other drugs. The two liquid preparations were subsequently extracted with ethyl acetate, pooled, then finally highly purified using a twin solid phase extraction on aminopropyl and silica cartridges. Residue was derivatized with MSTFA prior to injection. Analysis was conducted by gas chromatography coupled to a triple quadrupole mass spectrometer. Thirty cyclists were sampled and tested both in hair and in urine. Amphetamine was detected 10 times in hair (out of 19 analyses) compared to 6 times in urine (out of 30 analyses). Corticosteroids were detected 5 times in hair (methylprednisolone 1 case, triamcinolone acetonide 3 cases and hydrocortisone acetate 1 case) in hair (out of 12 analyses) compared to 12 times (triamcinolone acetonide 10 cases and betamethasone 2 cases) in urine (out of 30 analyses). Anabolic steroids were detected twice (nandrolone 1 case, and testosterone undecanoate 1 case) in hair (out of 25 analyses) compared to none in urine (out of 30 analyses).

  • gas chromatographic tandem mass spectrometric determination of anabolic steroids and their esters in hair application in doping control and meat quality control
    Journal of Chromatography B: Biomedical Sciences and Applications, 1999
    Co-Authors: Yvan Gaillard, Francois Vayssette, Antoine Balland, Gilbert Pepin
    Abstract:

    Abstract We have developed a powerful and simple sensitive method for testing hair for anabolic steroids and their esters. A 100-mg amount of powdered hair was treated with methanol in an ultrasonic bath for extraction of esters, then alkaline digested with 1 M NaOH for an Optimum Recovery of other drugs. The two liquid preparations were subsequently extracted with ethyl acetate, pooled, then finally highly purified using a twin solid-phase extraction on amino and silica cartridges. The residue was derivatized with N -methyl- N (trimethylsilyl)-trifluoracetamide (MSTFA) prior to injection. Analysis was conducted by gas chromatography coupled to a triple quadrupole mass spectrometer. The generally chosen parent ion was the molecular ion while two daughter ions were selected for each compound with collision energies ranging from −16 to −21 eV. Internal standards were nandrolone d 3 for non-esterified drugs and testosterone phenyl propionate for esters. The limits of detection calculated from an analysis of the blanks ( n =30) were 0.08 pg/mg for nandrolone, 6.20 pg/mg for boldenone, 0.07 pg/mg for methyl testosterone, 0.15 pg/mg for ethinyl estradiol, 2.10 pg/mg for metandienone, 0.86 pg/mg for testosterone propionate, 0.95 pg/mg for testosterone cypionate, 1.90 pg/mg for nandrolone decanoate, 3.10 pg/mg for testosterone decanoate and 4.80 pg/mg for testosterone undecanoate. Application to doping control has been demonstrated. In a series of 18 sportsmen, two tested positive for anabolic steroids in hair whereas urinalysis was negative for both of them. The first positive case was nandrolone and the second case concerned the identification of testosterone undecanoate. Measured in 10 white males aged between 22 and 31 years, the testosterone concentration was in the range 1.7–9.2 pg/mg (mean=5.0 pg/mg). The method was also applied in meat quality control. Of the 187 analyses realized based upon hair and urine sampling in slaughter houses, 23 were positive for anabolic steroids in hair: one case for boldenone, one case for metandienone, two cases for testosterone propionate, three cases for nandrolone, five cases for testosterone decanoate and 11 cases for methyl testosterone. In the meantime, urinalysis was always negative for these drugs or their metabolites.

Elham Heidari - One of the best experts on this subject based on the ideXlab platform.

  • extraction of epigallocatechin 3 gallate from green tea via supercritical fluid technology neural network modeling and response surface optimization
    Journal of Supercritical Fluids, 2013
    Co-Authors: S M Ghoreishi, Elham Heidari
    Abstract:

    Abstract In this study the extraction of (−)-Epigallocatechin-3-gallate (EGCG) from Iranian green tea was investigated by supercritical CO 2 with ethanol as co-solvent. Design of experiments and modeling were carried out with response surface methodology by Minitab software. The HPLC analysis of the extracted samples was used in conjunction with response surface design to optimize four operating variables of supercritical CO 2 extraction (pressure, temperature, CO 2 flow rate and extraction dynamic time). Optimum Recovery of EGCG (0.462 g/g) was obtained at 19.3 MPa, 43.7 °C, 106 min (dynamic) and 1.5 ml/min (CO 2 flow rate). Moreover, a three-layer artificial neural network was developed for modeling EGCG extraction from green tea. In this regard, different networks (by changing the number of neurons in the hidden layer and algorithm of network training) were compared with evaluation of networks accuracy in extraction Recovery prediction. Finally, the Levenberg–Marquardt algorithm with the six neurons in the hidden layer has been found to be the most suitable network.