Overlapped Data Processing Scheme for Accelerating Training and Validation in Machine Learning

Jinseo Choi

doi:10.1109/ACCESS.2022.3189373

2022-01-01

Jinseo Choi

Dong Hyun Kang *

DOI：10.1109/ACCESS.2022.3189373

原文链接

原文求助

摘要

En 中文

For several years, machine learning (ML) technologies open up new opportunities which solve traditional problems based on a rich set of hardware resources. Unfortunately, ML technologies sometimes waste available hardware resources (e.g., CPU and GPU) because they spend a lot of time waiting for a previous step inside ML procedure. In this paper, we first study data flows of the ML procedure in detail to find avoidable performance bottlenecks. Then, we propose ol.data, the first software-based data processing scheme that aims to (1) overlap training and validation steps in one epoch or two adjacent epochs, and (2) perform validation steps in parallel, which helps to significantly improve not only the computation time but also the resource utilization. To confirm the positive effectiveness of ol.data, we implemented a convolution neural network (CNN) model with ol.data and compared it with the traditional approaches, Numpy (i.e., baseline) and tf.data on three different datasets. As a result, we confirmed that ol.data reduces the inference time by up to 41.8% and increases the utilization of CPU and GPU resources by up to 75.7% and 38.7%, respectively.

Keyword:

Training

Graphics processing units

Task analysis

Data processing

Data models

Tensors

Hardware

Machine learning

TensorFlow

CPU

GPU utilization

overlapping

multiple threads

期刊

IEEE Access

IF:

3.6

论文数:

9.8W

被引数:

29.4W

机构

Changwon National University

学者数:

2.0K

论文数: 1.9K

被引数: 2

引用论文

Stability of some metal-ion complexes of tubercidin (= 7-deazaadenosine) in aqueous solution. An o-amino group inhibits complexation at N¹of purines!

J. Chem. Soc., Dalton Trans.

IF0

1991-01-01

PREAI

Liang-nian Ji; Nicolas A. Corfù; Helmut Sigel

Allele-Specific HLA Loss and Immune Escape in Lung Cancer Evolution

Cell

IF0

2017-11-01

OAAI

Nicholas McGranahan; Rachel Rosenthal; Crispin T. Hiley; Andrew J. Rowan; Thomas B.K. Watkins; Gareth A. Wilson; Nicolai J. Birkbak; Selvaraju Veeriah; Peter Van Loo; Javier Herrero; Charles Swanton; Charles Swanton; Mariam Jamal-Hanjani; Selvaraju Veeriah; Seema Shafi; Justyna Czyzewska-Khan; Diana Johnson; Joanne Laycock; Leticia Bosshard-Carter; Rachel Rosenthal; Pat Gorman; Robert E. Hynds; Gareth Wilson; Nicolai J. Birkbak; Thomas B.K. Watkins; Nicholas McGranahan; Stuart Horswell; Richard Mitter; Mickael Escudero; Aengus Stewart; Peter Van Loo; Andrew Rowan; Hang Xu; Samra Turajlic; Crispin Hiley; Christopher Abbosh; Jacki Goldman; Richard Kevin Stone; Tamara Denner; Nik Matthews; Greg Elgar; Sophia Ward; Marta Costa; Sharmin Begum; Ben Phillimore; Tim Chambers; Emma Nye; Sofia Graca; Maise Al Bakir; Kroopa Joshi; Andrew Furness; Assma Ben Aissa; Yien Ning Sophia Wong; Andy Georgiou; Sergio Quezada; John A. Hartley; Helen L. Lowe; Javier Herrero; David Lawrence; Martin Hayward; Nikolaos Panagiotopoulos; Shyam Kolvekar; Mary Falzon; Elaine Borg; Teresa Marafioti; Celia Simeon; Gemma Hector; Amy Smith; Marie Aranda; Marco Novelli; Dahmane Oukrif; Sam M. Janes; Ricky Thakrar; Martin Forster; Tanya Ahmad; Siow Ming Lee; Dionysis Papadatos-Pastos; Dawn Carnell; Ruheena Mendes; Jeremy George; Neal Navani; Asia Ahmed; Magali Taylor; Junaid Choudhary; Yvonne Summers; Raffaele Califano; Paul Taylor; Rajesh Shah; Piotr Krysiak; Kendadai Rammohan; Eustace Fontaine; Richard Booton; Matthew Evison; Phil Crosbie; Stuart Moss; Faiza Idries; Leena Joseph; Paul Bishop; Anshuman Chaturved; Anne Marie Quinn; Helen Doran; Angela Leek; Phil Harrison; Katrina Moore; Rachael Waddington; Juliette Novasio; Fiona Blackhall; Jane Rogan; Elaine Smith; Caroline Dive; Jonathan Tugwood; Ged Brady; Dominic G. Rothwell; Francesca Chemi; Jackie Pierce; Sakshi Gulati; Babu Naidu; Gerald Langman; Simon Trotter; Mary Bellamy; Hollie Bancroft; Amy Kerr; Salma Kadiri; Joanne Webb; Gary Middleton; Madava Djearaman; Dean Fennell; Jacqui A. Shaw; John Le Quesne; David Moore; Apostolos Nakas; Sridhar Rathinam; William Monteiro; Hilary Marshall; Louise Nelson; Jonathan Bennett; Joan Riley; Lindsay Primrose; Luke Martinson; Girija Anand; Sajid Khan; Anita Amadi; Marianne Nicolson; Keith Kerr; Shirley Palmer; Hardy Remmen; Joy Miller; Keith Buchan; Mahendran Chetty; Lesley Gomersall; Jason Lester; Alison Edwards; Fiona Morgan; Haydn Adams; Helen Davies; Malgorzata Kornaszewska; Richard Attanoos; Sara Lock; Azmina Verjee; Mairead MacKenzie; Maggie Wilcox; Harriet Bell; Allan Hackshaw; Yenting Ngai; Sean Smith; Nicole Gower; Christian Ottensmeier; Serena Chee; Benjamin Johnson; Aiman Alzetani; Emily Shaw; Eric Lim; Paulo De Sousa; Monica Tavares Barbosa; Alex Bowman; Simon Jordan; Alexandra Rice; Hilgardt Raubenheimer; Chiara Proli; Maria Elena Cufari; John Carlo Ronquillo; Angela Kwayie; Harshil Bhayani; Morag Hamilton; Yusura Bakar; Natalie Mensah; Lyn Ambrose; Anand Devaraj; Silviu Buderi; Jonathan Finch; Leire Azcarate; Hema Chavan; Sophie Green; Hillaria Mashinga; Andrew G. Nicholson; Kelvin Lau; Michael Sheaff; Peter Schmid; John Conibear; Veni Ezhil; Babikir Ismail; Melanie Irvin-sellers; Vineet Prakash; Peter Russell; Teresa Light; Tracey Horey; Sarah Danson; Jonathan Bury; John Edwards; Jennifer Hill; Sue Matthews; Yota Kitsanta; Kim Suvarna; Patricia Fisher; Allah Dino Keerio; Michael Shackcloth; John Gosney; Pieter Postmus; Sarah Feeney; Julius Asante-Siaw; Hugo J.W.L. Aerts; Stefan Dentro; Christophe Dessimoz

First Vehicle Arrival Time Prediction at Signalized Intersection Based on Wavelet-Elman Neural Network

NeuroQuantology

IF0

2018-06-25

PREAI

Guchang Ao; Huiling Zhang; Linyu Yang; Wuting Jiang

Gradient-based learning applied to document recognition 基于梯度的学习在文档识别中的应用

PROCEEDINGS OF THE IEEE

IF25.9

1998-01-01

3.8W

PREAI