Assessing hyper parameter optimization and speedup for convolutional neural networks

Sajid Nazir, Shushma Patel, Dilip Patel

Research output: Contribution to journalArticle

26 Downloads (Pure)

Abstract

The increased processing power of Graphical Processing Units (GPUs) and the availability of large image datasets has fostered a renewed interest in extracting semantic information from images. Promising results for complex image categorization problems have been achieved using deep learning, with neural networks comprising of many layers. Convolutional Neural Network (CNN) is one such architecture providing more opportunities for image classification. Advances in CNN enable the development of training models using large labelled image datasets, but the hyper parameters need to be specified, which is challenging and complex, due to the large number of parameters. A substantial amount of computational power and processing time is required to determine the optimal hyper parameters to define a model yielding good results. This paper provides a survey of the hyper parameter search and optimization methods for CNN architectures.
Original languageEnglish
JournalInternational Journal of Artificial Intelligence and Machine Learning (IJAIML)
Publication statusAccepted/In press - 18 Jan 2020

Fingerprint

Neural networks
Processing
Image classification
Network architecture
Semantics
Availability
Deep learning

Keywords

  • Deep Learning, Artificial Intelligence, Cognitive Image Processing, Object recognition, Convolution, Semantics, Machine learning, Hidden Layers

Cite this

Nazir, S., Patel, S., & Patel, D. (Accepted/In press). Assessing hyper parameter optimization and speedup for convolutional neural networks. International Journal of Artificial Intelligence and Machine Learning (IJAIML).
Nazir, Sajid ; Patel, Shushma ; Patel, Dilip. / Assessing hyper parameter optimization and speedup for convolutional neural networks. In: International Journal of Artificial Intelligence and Machine Learning (IJAIML). 2020.
@article{1cd4b82514fe4b1a840d0220020273e5,
title = "Assessing hyper parameter optimization and speedup for convolutional neural networks",
abstract = "The increased processing power of Graphical Processing Units (GPUs) and the availability of large image datasets has fostered a renewed interest in extracting semantic information from images. Promising results for complex image categorization problems have been achieved using deep learning, with neural networks comprising of many layers. Convolutional Neural Network (CNN) is one such architecture providing more opportunities for image classification. Advances in CNN enable the development of training models using large labelled image datasets, but the hyper parameters need to be specified, which is challenging and complex, due to the large number of parameters. A substantial amount of computational power and processing time is required to determine the optimal hyper parameters to define a model yielding good results. This paper provides a survey of the hyper parameter search and optimization methods for CNN architectures.",
keywords = "Deep Learning, Artificial Intelligence, Cognitive Image Processing, Object recognition, Convolution, Semantics, Machine learning, Hidden Layers",
author = "Sajid Nazir and Shushma Patel and Dilip Patel",
note = "Acceptance in SAN (note from author) AAM: unknown publisher policy - made file open and contacted publisher 27/1/20 DC - Upon publication, remove AAM with VoR and update with rights statement provided by publisher. 31/1/20 DC",
year = "2020",
month = "1",
day = "18",
language = "English",

}

Nazir, S, Patel, S & Patel, D 2020, 'Assessing hyper parameter optimization and speedup for convolutional neural networks', International Journal of Artificial Intelligence and Machine Learning (IJAIML).

Assessing hyper parameter optimization and speedup for convolutional neural networks. / Nazir, Sajid; Patel, Shushma; Patel, Dilip.

In: International Journal of Artificial Intelligence and Machine Learning (IJAIML), 18.01.2020.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Assessing hyper parameter optimization and speedup for convolutional neural networks

AU - Nazir, Sajid

AU - Patel, Shushma

AU - Patel, Dilip

N1 - Acceptance in SAN (note from author) AAM: unknown publisher policy - made file open and contacted publisher 27/1/20 DC - Upon publication, remove AAM with VoR and update with rights statement provided by publisher. 31/1/20 DC

PY - 2020/1/18

Y1 - 2020/1/18

N2 - The increased processing power of Graphical Processing Units (GPUs) and the availability of large image datasets has fostered a renewed interest in extracting semantic information from images. Promising results for complex image categorization problems have been achieved using deep learning, with neural networks comprising of many layers. Convolutional Neural Network (CNN) is one such architecture providing more opportunities for image classification. Advances in CNN enable the development of training models using large labelled image datasets, but the hyper parameters need to be specified, which is challenging and complex, due to the large number of parameters. A substantial amount of computational power and processing time is required to determine the optimal hyper parameters to define a model yielding good results. This paper provides a survey of the hyper parameter search and optimization methods for CNN architectures.

AB - The increased processing power of Graphical Processing Units (GPUs) and the availability of large image datasets has fostered a renewed interest in extracting semantic information from images. Promising results for complex image categorization problems have been achieved using deep learning, with neural networks comprising of many layers. Convolutional Neural Network (CNN) is one such architecture providing more opportunities for image classification. Advances in CNN enable the development of training models using large labelled image datasets, but the hyper parameters need to be specified, which is challenging and complex, due to the large number of parameters. A substantial amount of computational power and processing time is required to determine the optimal hyper parameters to define a model yielding good results. This paper provides a survey of the hyper parameter search and optimization methods for CNN architectures.

KW - Deep Learning, Artificial Intelligence, Cognitive Image Processing, Object recognition, Convolution, Semantics, Machine learning, Hidden Layers

M3 - Article

ER -

Nazir S, Patel S, Patel D. Assessing hyper parameter optimization and speedup for convolutional neural networks. International Journal of Artificial Intelligence and Machine Learning (IJAIML). 2020 Jan 18.