A feasibility study into the acceleration of multivariate analysis techniques using Graphics Processing Units (GPUs) will be presented. The MLP-based Artificial Neural Network method contained in the ...TMVA framework has been chosen as a focus for investigation. It was found that the network training time on a GPU was lower than for CPU execution as the complexity of the network was increased. In addition, multiple neural networks can be trained simultaneously on a GPU within the same time taken for single network training on a CPU. This could be potentially leveraged to provide a qualitative performance gain in data classification.
In high-energy physics, with the search for ever smaller signals in ever
larger data sets, it has become essential to extract a maximum of the available
information from the data. Multivariate ...classification methods based on machine
learning techniques have become a fundamental ingredient to most analyses. Also
the multivariate classifiers themselves have significantly evolved in recent
years. Statisticians have found new ways to tune and to combine classifiers to
further gain in performance. Integrated into the analysis framework ROOT, TMVA
is a toolkit which hosts a large variety of multivariate classification
algorithms. Training, testing, performance evaluation and application of all
available classifiers is carried out simultaneously via user-friendly
interfaces. With version 4, TMVA has been extended to multivariate regression
of a real-valued target vector. Regression is invoked through the same user
interfaces as classification. TMVA 4 also features more flexible data handling
allowing one to arbitrarily form combined MVA methods. A generalised boosting
method is the first realisation benefiting from the new framework.
In high-energy physics, with the search for ever smaller signals in ever larger data sets, it has become essential to extract a maximum of the available information from the data. Multivariate ...classification methods based on machine learning techniques have become a fundamental ingredient to most analyses. Also the multivariate classifiers themselves have significantly evolved in recent years. Statisticians have found new ways to tune and to combine classifiers to further gain in performance. Integrated into the analysis framework ROOT, TMVA is a toolkit which hosts a large variety of multivariate classification algorithms. Training, testing, performance evaluation and application of all available classifiers is carried out simultaneously via user-friendly interfaces. With version 4, TMVA has been extended to multivariate regression of a real-valued target vector. Regression is invoked through the same user interfaces as classification. TMVA 4 also features more flexible data handling allowing one to arbitrarily form combined MVA methods. A generalised boosting method is the first realisation benefiting from the new framework.
The nominal operating cell temperature (NOCT) was developed as a reference characterization test procedure to quantify the module cell temperature for different module designs in a standard reference ...environment (SRE) of 20 °C ambient temperature, 800 W/m 2 irradiance and 1 m/s wind speed 1. The NOCT value is a key performance parameter to be measured as per the IEC 61215 (Edition 2) standard, and it is required to be independently measured and reported to qualify for the incentive programs of various agencies including California Energy Commission. Ideally, the NOCT value of a specific module design should be identical irrespective of testing laboratory, location, month or season. The objectives of this NOCT Round-Robin testing were: (i) to identify if NOCT values are significantly influenced by the testing approach or site specific test conditions experienced by different labs in the world (referred to as reproducibility); (ii) to identify if NOCT values are dependent on the month or season within any single laboratory (referred to as repeatability), and (iii) to identify if NOCT values are significantly influenced by the type (thermocouple or RTD) and position (backsheet or cell) of thermal sensor used by different labs in the world. A total of eight polycrystalline silicon modules with a nameplate rating of 217 watts were continuously tested over a year in eight different test laboratories around the world, one at each participating laboratory. All test samples are of the same model, were supplied by a single manufacturer, were received as a batch and were assumed to be identical with no variability in manufacturing. The test laboratories were not supplied with any specific testing procedure and they were asked to perform the testing as per their established standard operating procedures (SOP).