Speaker
Description
Highly precise simulations of elementary particles interaction and processes are fundamental to accurately reproduce and interpret the experimental results in High Energy Physics (HEP) detectors and to correctly reconstruct the particle flows. Today, detector simulations typically rely on Monte Carlo-based methods which are extremely demanding in terms of computing resources. The need for simulated data at future experiments - like the ones that will run at the High Luminosity Large Hadron Collider (HL-LHC) - are expected to increase by orders of magnitude, increasing drastically the computational challenge. This expectation motivates the research for alternative deep learning-based simulation strategies. In this research we speed-up HEP detector simulations for the specific case of calorimeters using Generative Adversarial Networks (GANs) with a huge factor of over 150 000x compared to the standard Monte Carlo simulations. This could only be achieved by designing smart convolutional 2D network architectures for generating 3D images representing the detector volume. Detailed physics evaluation shows an accuracy similar to the Monte Carlo simulation.
Furthermore, we quantize the data format for the neural network architecture (float32) with the Intel Low Precision Optimization tool (LPOT) to a reduced precision (int8) data format. This results in an additional $1.8$x speed-up on modern Intel hardware while maintaining the physics accuracy. These excellent results consolidate the beneficial use of GANs for future fast detector simulations.
