IIUM Repository

A multilayered convolutional sparse coding framework for modeling of pooling operation of convolution neural networks

, Abdul Wahid and Khan, Adnan Umar and , Mukhtarullah and Khan, Sheroz and Shah, Jawad (2019) A multilayered convolutional sparse coding framework for modeling of pooling operation of convolution neural networks. In: 2019 IEEE 6th International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA 2019), 27 - 29 Aug 2019, Kuala Lumpur, Malaysia.

[img] PDF - Published Version
Restricted to Repository staff only

Download (1MB) | Request a copy
[img] PDF (SCOPUS) - Supplemental Material
Restricted to Repository staff only

Download (245kB) | Request a copy

Abstract

Convolutional Sparse Coding (CSC) framework has been proposed recently to explain relation between Convolutional Neural Networks (CNNs) and sparse coding theory. The multilayered version of CSC(ML-CSC) is shown to be connected to forward pass of CNNs and dictionary learning and sparse coding algorithms of this model are analyzed for solving classification and inverse problems in image processing. However open problems like effect of pooling operations, batch normalization and dictionary learning in context of ML-CSC framework remain challenging issues especially in implementation scenarios. In this work we implement the framework for multi layered version of CSC with incorporation of pooling operations applied on real images and analyze the performance of resulting model. We demonstrate that with optimal parameters selected for sparsity of feature maps, the pooling operation (here max pooling) when used layered wise in ML-CSC framework improves the effective dictionaries and resulting feature maps, which in turn improves the reconstruction accuracy of images after multilayered implementation.

Item Type: Conference or Workshop Item (Plenary Papers)
Additional Information: 3930/80454
Uncontrolled Keywords: Sparse coding, convolutional neural networks, pooling operation,multi layer convolutional sparse coding
Subjects: T Technology > T Technology (General)
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button): Kulliyyah of Engineering
Kulliyyah of Engineering > Department of Electrical and Computer Engineering
Depositing User: dr Sheroz Khan
Date Deposited: 13 May 2020 11:03
Last Modified: 13 Jul 2020 14:50
URI: http://irep.iium.edu.my/id/eprint/80454

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year