Corrigendum: Applications and techniques for fast machine learning in science.

[This corrects the article DOI: 10.3389/fdata.2022.787421.].

Applications and Techniques for Fast Machine Learning in Science.

In this community review report, we discuss applications and techniques for fast machine learning (ML) in science-the concept of integrating powerful ML methods into the real-time experimental data processing loop to accelerate scientific discovery. The material for the report builds on two workshops held by the Fast ML for Science community and covers three main areas: applications for fast ML across a number of scientific domains; techniques for training and implementing performant and resource-efficient ML algorithms; and computing architectures, platforms, and technologies for deploying these algorithms. We also present overlapping challenges across the multiple scientific domains where common solutions can be found. This community report is intended to give plenty of examples and inspiration for scientific discovery through integrated and accelerated ML solutions. This is followed by a high-level overview and organization of technical advances, including an abundance of pointers to source material, which can enable these breakthroughs.

The accurate prediction of sweat rate from energy expenditure and air temperature: a proof-of-concept study.

This proof-of-concept study used a web application to predict runner sweat losses using only energy expenditure and air temperature. A field study (FS) of n = 37 runners was completed with n = 40 sweat loss observations measured over 1 h (sweat rate, SR). Predictions were also compared with 10 open literature (OL) studies in which individual runner SR was reported (n = 82; 109 observations). Three prediction accuracy metrics were used: for FS, the mean absolute error (MAE) and concordance correlation coefficient (CCC) were calculated to include a 95% confidence interval [CI]; for OL, the percentage concordance (PC) was examined against calculation of accumulated under- and over-drinking potential. The MAE for FS runners was 0.141 kg [0.105, 0.177], which was less than estimated scale weighing error on 85% of occasions. The CCC was 0.88 [0.82, 0.93]. The PC for OL was 96% for avoidance of both under- and over-drinking and 93% overall. All accuracy metrics and their CIs were below acceptable error tolerance. Input errors of ±10% and ±1 °C for energy expenditure and air temperature dropped the PC to between 84% and 90%. This study demonstrates the feasibility of accurately predicting SR from energy expenditure and air temperature alone. Novelty Results demonstrate that accurate runner SR prediction is possible with knowledge of only energy expenditure and air temperature. SR prediction error was smaller than scale weighing error in 85% of observations. Accurate runner SR prediction could help mitigate the common risks of over- and under-drinking.

Distance-Weighted Graph Neural Networks on FPGAs for Real-Time Particle Reconstruction in High Energy Physics.

Graph neural networks have been shown to achieve excellent performance for several crucial tasks in particle physics, such as charged particle tracking, jet tagging, and clustering. An important domain for the application of these networks is the FGPA-based first layer of real-time data filtering at the CERN Large Hadron Collider, which has strict latency and resource constraints. We discuss how to design distance-weighted graph networks that can be executed with a latency of less than one µs on an FPGA. To do so, we consider a representative task associated to particle reconstruction and identification in a next-generation calorimeter operating at a particle collider. We use a graph network architecture developed for such purposes, and apply additional simplifications to match the computing constraints of Level-1 trigger systems, including weight quantization. Using the hls4ml library, we convert the compressed models into firmware to be implemented on an FPGA. Performance of the synthesized models is presented both in terms of inference accuracy and resource usage.