Reduced Precision Strategies for Deep Learning : A High Energy Physics Generative Adversarial Network Use Case

RWTH Publications (RWTH Aachen)

Abstract

Deep learning is finding its way into high energy physics by replacing traditional Monte Carlo simulations.However, deep learning still requires an excessive amount of computational resources.A promising approach to make deep learning more efficient is to quantize the parameters of the neural networks to reduced precision.Reduced precision computing is extensively used in modern deep learning and results to lower execution inference time, smaller memory footprint and less memory bandwidth.In this paper we analyse the effects of low precision inference on a complex deep generative adversarial network model.The use case which we are addressing is calorimeter detector simulations of subatomic particle interactions in accelerator based high energy physics.We employ the novel Intel low precision optimization tool (iLoT) for quantization and compare the results to the quantized model from TensorFlow Lite.In the performance benchmark we gain a speed-up of 1.73x on Intel hardware for the quantized iLoT model compared to the initial, not quantized, model.With different physics-inspired self-developed metrics, we validate that the quantized iLoT model shows a lower loss of physical accuracy in comparison to the TensorFlow Lite model.

Authors 8

1. Florian Rehm Aachen

   European Organization for Nuclear Research · RWTH Aachen University

   Affiliation as printed

   CERN, Switzerland, --- Select a Country ---

   RWTH Aachen University, Germany, --- Select a Country ---

2. Sofia Vallecorsa

   European Organization for Nuclear Research

   Affiliation as printed

   CERN, Switzerland, --- Select a Country ---

3. Vikram A. Saletore

   Intel (United States)

   Affiliation as printed

   Intel, U.S.A., --- Select a Country ---

4. Hans Pabst

   Intel (United States)

   Affiliation as printed

   Intel, U.S.A., --- Select a Country ---

5. Adel Chaibi

   Intel (United States)

   Affiliation as printed

   Intel, U.S.A., --- Select a Country ---

6. Valeriu Codreanu

   SURFsara (Netherlands)

   Affiliation as printed

   SURFsara, The Netherlands, --- Select a Country ---

7. K. Borras Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University, Germany, --- Select a Country ---

8. Dirk Krücker

   Deutsches Elektronen-Synchrotron DESY

   Affiliation as printed

   DESY, Germany, --- Select a Country ---