In this paper, we propose a new directional multi-resolution ridgelet network (DMRN) based on the ridgelet frame theory, which uses the ridgelet as the activation function in a hidden layer. For the multi-resolution properties of the ridgelet function in the direction besides scale and position, DMRN has great capabilities in catching essential features of direction-rich data. It proves to be able to approximate any multivariate function in a more stable and efficient way, and optimal in approximating functions with spatial inhomogeneities. Besides, using binary ridgelet frame as the mathematical foundation in its construction, DMRN is more flexible with a simple structure. The construction and the learning algorithm of DMRN are given. Its approximation capacity and approximation rate are also analyzed in detail. Possibilities of applications to regression and recognition are included to demonstrate its superiority to other methods and feasibility in practice. Both theory analysis and simulation results prove its high efficiency.
YANG Shuyuan, JIAO Licheng, WANG Min
. A new directional multi-resolution ridgelet network[J]. Frontiers of Electrical and Electronic Engineering, 2008
, 3(2)
: 198
-203
.
DOI: 10.1007/s11460-008-0026-2
1. McCulloch W S Pitts W A logical calculus of the ideasimmanent in nervous activityBulletin ofMathematical Biophysics 1943 5115133. doi:10.1007/BF02478259
2. Park J Sandberg I W Universal approximation usingradial-basis-function networksNeural Computation 1991 3(2)246257. doi:10.1162/neco.1991.3.2.246
3. Zhang Q Benveniste A Wavelet networksIEEE Transactions on Neural Networks 1992 3(6)889898. doi:10.1109/72.165591
4. Hubel D H Wiesel T N Receptive fields, binocularinteraction and functional architecture in the cat's visual cortexThe Journal of Physiology 1962 160106154
5. Candàs E J Harmonicanalysis of neural networksApplied andComputational Harmonic Analysis 1999 6(2)197218. doi:10.1006/acha.1998.0248
6. Donoho D L Tightframes of k-plane ridgelets andthe problem of representing objects that are smooth away from d-dimensionalsingularitiesIn: Proceedings of the NationalAcademy of Sciences, USA 1999 96(5)18281833. doi:10.1073/pnas.96.5.1828
7. Candàs E J Donoho D L Ridgelets: a key to higher-dimensionalintermittencyPhilosophical Transactionsof the Royal Society of London Series A 1999 357(1760)24952509. doi:10.1098/rsta.1999.0444
8. Starck J L Candàs E J Donoho D L The curvelet transform for image denoisingIEEE Transactions on Image Processing 2002 11(6)670684. doi:10.1109/TIP.2002.1014998
9. Grochenig K Accelerationof the Frame AlgorithmIEEE Transactionson Signal Processing 1993 41(12)33313340. doi:10.1109/78.258077
10. Candàs E J Ridgelet:theory and applicationsDissertation forthe Doctoral Degree. CA: StanfordUniversity 1998
11. Daubechies I Thewavelet transform: time-frequency localization and signal analysisIEEE Transactions on Information Theory 1990 36(5)9611005. doi:10.1109/18.57199
12. Donoho D L Orthonormalridgelets and linear singularitiesSIAMJournal on Mathematical Analysis 2000 31(5)10621099. doi:10.1137/S0036141098344403
13. Do M N Vetterli M The finite ridgelet transformfor image representationIEEE Transactionson Image Processing 2003 12(1)1628. doi:10.1109/TIP.2002.806252
14. Lin W Kovvali N Carin L Ridgelet-based implementation of multi-resolution timedomainIEEE Transactions on Antennas andPropagation 2005 53(8)26882699. doi:10.1109/TAP.2005.851756
15. Hou B Liu F Jiao L C Linear feature detection based on ridgeletScience in China (Series E) 2003 46(2)141152. doi:10.1360/03ye9015
16. Yang S Y Jiao L C Wang M An adaptive ridgelet neural network modelJournal of Xidian University 2005 32(6)890894(in Chinese)
