Speaker
Dr
Alfio Lazzaro
(Universita degli Studi di Milano & INFN, Milano)
Description
With the startup of the LHC experiments, the community will be focused on the data analysis of the collected data. The complexity of the data analyses will be a key factor to find eventual new phenomena. For such a reason many data analysis tools are being developed in the last years. allowing the use of different techniques, such as likelihood-based procedures, neural networks, boost decision trees. In particular the likelihood-based procedures allow the estimation of unknown parameters based on a given input sample. Complex likelihood functions, with several free parameters, many independent variables and large data sample, can be very CPU-time consuming for their calculation. Furthermore for a good estimation it is required the generation of several simulated samples of events from the probability density functions, so the whole procedure which can be CPU-time consuming. In this presentation I will show how the likelihood calculation, the normalization integrals calculation, and the events generation can be parallelized using MPI techniques to scale over multiple nodes or multi-threads for multi-cores in a single node. We will present the speed-up improvements obtained in typical physics applications such as complex maximum likelihood fits using the RooFit and RooStats packages. We will also show results of hybrid parallelization between MPI and multi-threads, to take full advantage of multi-core architectures.
Primary author
Dr
Alfio Lazzaro
(Universita degli Studi di Milano & INFN, Milano)
