Radiotekhnika
Publishing house Radiotekhnika

"Publishing house Radiotekhnika":
scientific and technical literature.
Books and journals of publishing houses: IPRZHR, RS-PRESS, SCIENCE-PRESS


Тел.: +7 (495) 625-9241

 

Classification and generation of images with a hierarchically connected neural network

Keywords:

Diane Sekou Abdel Kader – Post-graduate Student, Assistant, HVE Department of Control Engineering, Moscow State Institute of Radio Engineering, Electronics and Automation. E-mail: sekoudiane1990@gmail.com


The paper introduces a method for construction and learning of neural networks which is based on organization of hierarchical connectivity between neurons of feed-forward neural network and on the use of biologically inspired principles for neuronal selectivity formation. Proposed approach allows detection of meaningful features in input information image. Complexity of the features grows from lower layers of neural network to the upper ones. Detected features can be used to increase the quality of image classification. Moreover image generation is possible based on the set of high-level features via multi-stage decoding of their values. Results of modeling of neural network with hierarchical connectivity are presented for tasks of recognition and generation of handwritten symbol images. Finally application perspectives of the proposed approach are discussed in the some tasks of data mining and intellectual control.
References:

 

  1. Hawkins J., Blakeslee S. On Intelligence. USA: Henry Holt and Company. 2004. 261 p.
  2. Umryuxin E.A. Problemy' modelirovaniya funkczij mozga (nejrokiberneticheskij podxod) // Nejrokomp'yutery': razrabotka, primenenie. 2013. № 8. S. 54–59.
  3. Chernavskij D.S., Karp V.P., Nikitin A.P., Chernavskaya O.D. Odin iz variantov konstrukczii apparata my'shleniya i model' obucheniya rechi // Sbornik trudov XV Vseross. nauchno-texnich. konf. «Nejroinformatika-2013». V 3-x chastyax. Ch. 1. M.: Izd-vo NIJaU MIFI. 2013. 240 s.
  4. Salakhutdinov R., Hinton G. Deep Boltzmann Machines // Proceedings of the 12th International Conference on Artificial Intelligence and Statistics (AISTATS). 2009. Clearwater Beach. Florida, USA. V. 5 of JMLR.
  5. Ciresan D.C. et al. Flexible, High Performance Convolutional Neural Networks for Image Classification // Twenty-Second International Joint Conference on Artificial Intelligence. 2011.
  6. Ananthanarayanan R., Esser S., Simon H., Modha D. The Cat is Out of the Bag: Cortical Simulations with 109 Neurons, 1013 Synapses // SC09. November 14-20 2009. Portland, Oregon, USA.
  7. Bolotova Ju.A., Kermani A.K., Spiczy'n V.G. Raspoznavanie simvolov na czvetnom fone na osnove ierarxicheskoj vremennoj modeli s predobrabotkoj fil'trami Gabora // E'lektromagnitny'e volny' i e'lektronny'e sistemy'. 2012. № 1. S. 14-19.
  8. Bienenstock E., Cooper L., and Munro P.W. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. // The Journal of Neuroscience. January 1982. № 1. R. 32-48.
  9. LeCun Y., Bottou L., Bengio Y., and Haffner P. Gradient-based learning applied to document recognition // Proceedings of the IEEE. November 1998.
  10. Ciresan D., Meier U., Schmidhuber J. Multi-column Deep Neural Networks for Image Classification. Technical Report No. IDSIA-04-12, [E'lektronny'j resurs] // Rezhim dostupa: http://arxiv.org/pdf/1202.2745v1.pdf

 

© Издательство «РАДИОТЕХНИКА», 2004-2017            Тел.: (495) 625-9241                   Designed by [SWAP]Studio