A novel hybrid approach for combining deep and traditional neural networks

Rui Zhang*, Shufei Zhang, Kaizhu Huang

*Corresponding author for this work

Research output: Chapter in Book or Report/Conference proceedingConference Proceedingpeer-review

Abstract

Over last fifty years, Neural Networks (NN) have been important and active models in machine learning and pattern recognition. Among different types of NNs, Back Propagation (BP) NN is one popular model, widely exploited in various applications. Recently, NNs attract even more attention in the community because a deep learning structure (if appropriately adopted) could significantly improve the learning performance. In this paper, based on a probabilistic assumption over the output neurons, we propose a hybrid strategy that manages to combine one typical deep NN, i.e., Convolutional NN (CNN) with the popular BP. We present the justification and describe the detailed learning formulations. A series of experiments validate that the hybrid approach could largely improve the accuracy for both CNN and BP on two largescale benchmark data sets, i.e., MNIST and USPS. In particular, the proposed hybrid method significantly reduced the error rates of CNN and BP respectively by 11.72% and 28.89% on MNIST.

Original languageEnglish
Title of host publicationNeural Information Processing - 21st International Conference, ICONIP 2014, Proceedings
EditorsChu Kiong Loo, Keem Siah Yap, Kok Wai Wong, Andrew Teoh, Kaizhu Huang
PublisherSpringer Verlag
Pages349-356
Number of pages8
ISBN (Electronic)9783319126425
DOIs
Publication statusPublished - 2014
Event21st International Conference on Neural Information Processing, ICONIP 2014 - Kuching, Malaysia
Duration: 3 Nov 20146 Nov 2014

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8836
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st International Conference on Neural Information Processing, ICONIP 2014
Country/TerritoryMalaysia
CityKuching
Period3/11/146/11/14

Fingerprint

Dive into the research topics of 'A novel hybrid approach for combining deep and traditional neural networks'. Together they form a unique fingerprint.

Cite this