Stochastic diagonal approximate greatest descent in convolutional neural networks

© 2017 IEEE. Deep structured of Convolutional Neural Networks (CNN) has recently gained intense attention in development due to its good performance in object recognition. One of the crucial components in CNN is the learning mechanism of weight parameters through backpropagation. In this paper, stoc...

Full description

Bibliographic Details
Main Authors: Tan, H., Lim, Hann, Harno, H.
Format: Conference Paper
Published: 2017
Online Access:http://hdl.handle.net/20.500.11937/65863
_version_ 1848761220368171008
author Tan, H.
Lim, Hann
Harno, H.
author_facet Tan, H.
Lim, Hann
Harno, H.
author_sort Tan, H.
building Curtin Institutional Repository
collection Online Access
description © 2017 IEEE. Deep structured of Convolutional Neural Networks (CNN) has recently gained intense attention in development due to its good performance in object recognition. One of the crucial components in CNN is the learning mechanism of weight parameters through backpropagation. In this paper, stochastic diagonal Approximate Greatest Descent (SDAGD) is proposed to train weight parameters in CNN. SDAGD adopts the concept of multistage control system and diagonal Hessian approximation for weight optimization. It can be defined into two-phase optimization. In phase 1, when an initial guessing point is far from the solution, SDAGD constructs local search regions to determine the step length of next iteration at the boundary of search region. Subsequently, when the solution is at the final search region, SDAGD will shift to phase 2 by approximating Newton method to obtain a fast weight convergence. The calculation of Hessian in diagonal approximation results in less computational cost as compared to full Hessian calculation. The experiment showed that SDAGD learning algorithm could achieve misclassification rate of 8.85% on MNIST dataset.
first_indexed 2025-11-14T10:28:12Z
format Conference Paper
id curtin-20.500.11937-65863
institution Curtin University Malaysia
institution_category Local University
last_indexed 2025-11-14T10:28:12Z
publishDate 2017
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-658632018-02-19T08:06:52Z Stochastic diagonal approximate greatest descent in convolutional neural networks Tan, H. Lim, Hann Harno, H. © 2017 IEEE. Deep structured of Convolutional Neural Networks (CNN) has recently gained intense attention in development due to its good performance in object recognition. One of the crucial components in CNN is the learning mechanism of weight parameters through backpropagation. In this paper, stochastic diagonal Approximate Greatest Descent (SDAGD) is proposed to train weight parameters in CNN. SDAGD adopts the concept of multistage control system and diagonal Hessian approximation for weight optimization. It can be defined into two-phase optimization. In phase 1, when an initial guessing point is far from the solution, SDAGD constructs local search regions to determine the step length of next iteration at the boundary of search region. Subsequently, when the solution is at the final search region, SDAGD will shift to phase 2 by approximating Newton method to obtain a fast weight convergence. The calculation of Hessian in diagonal approximation results in less computational cost as compared to full Hessian calculation. The experiment showed that SDAGD learning algorithm could achieve misclassification rate of 8.85% on MNIST dataset. 2017 Conference Paper http://hdl.handle.net/20.500.11937/65863 10.1109/ICSIPA.2017.8120653 restricted
spellingShingle Tan, H.
Lim, Hann
Harno, H.
Stochastic diagonal approximate greatest descent in convolutional neural networks
title Stochastic diagonal approximate greatest descent in convolutional neural networks
title_full Stochastic diagonal approximate greatest descent in convolutional neural networks
title_fullStr Stochastic diagonal approximate greatest descent in convolutional neural networks
title_full_unstemmed Stochastic diagonal approximate greatest descent in convolutional neural networks
title_short Stochastic diagonal approximate greatest descent in convolutional neural networks
title_sort stochastic diagonal approximate greatest descent in convolutional neural networks
url http://hdl.handle.net/20.500.11937/65863