Speakers
Description
ML/DL techniques have shown their power in the improvement of several studies and tasks in HEP, in particular, especially in physics analysis. Our approach has been to take a number of the ML/DL tools provided by several Open Source platforms applying them to several classification problems, for instance, to the ttbar resonance extraction in LHC experiment.
A comparison has been made between RStudio and a Python-based environment. Boosted Decision Trees, Random Forest, ANN, etc have been used and optimized by using hyperparameters to control overfitting. The execution CPU times for each choice were collected in order to obtain a complete picture of the performance.
On top of this, Data simulation with traditional models is computationally very demanding, making the use of generative models an alternative for generating simulated Monte Carlo events with a similar quality at a lower cost. This could help to have at our disposal a larger statistic of simulated data for a better sensitivity and a more precise evaluation of systematic errors in possible discoveries of Physics Beyond the Standard Model.
In this work, we study the use of generative models based on Deep Learning as faster Monte Carlo event generators in the LHC context, reducing the time and energy cost of currently used methods. In particular, we focus on different variants of Variational Autoencoders, taking as a starting point the already known beta-VAE and proposing a new variant: the alpha-VAE, that improves the results in some experiments. In fact, the alpha-VAE is a simplification of the widely used beta-VAE.
Considerations will be made about the reliability of these simulated data when they are produced with very high statistics.
| Consider for long presentation | Yes |
|---|
