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

TL;DR

This paper explores low precision quantization of deep generative adversarial networks for high energy physics simulations, demonstrating significant speed-up and maintained accuracy using novel and existing quantization tools.

Contribution

It introduces the use of the Intel iLoT tool for quantizing complex GANs in physics, showing improved performance and accuracy preservation over TensorFlow Lite.

Findings

1.73x speed-up on Intel hardware with iLoT quantization

Lower physical accuracy loss with iLoT compared to TensorFlow Lite

Effective quantization of complex physics-based GAN models

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.