An Interruptible Connectionist Model for Real-Time Pattern Recognition
We describe an approach used for the conception of neural real-time pattern recognition systems which are interruptible, that is, able to give answers before the computing completion. This method is based on dynamic coding of information in the neural network: the proposed neuron model performs a time integration of its inputs and emits binary spikes trains. We have designed a dynamic shared-weights multi-layer neural classifier with recognition rate close to a more classical pattern recognition network, suitable for real-time applications with early hypothesis production. Because of its low computing complexity, this model seems to be well suited to hardware implementations.
Unable to display preview. Download preview PDF.
- Rebourg J.L., Muller J.D., Samuelides M. SPIKE_4096: A neural integrated circuit for image segmentation. Submitted to ICES’98, International Conference on Evolvable Systems: From Biology to Hardware, Lausanne, Switzerland, 23–26 sept. 1998.Google Scholar
- Thorpe S.J. Spike arrival times: A highly efficient coding scheme for neural networks. In Eckmiller, Hartman, and Hauske (Eds.,), Parallel Processing in Neural Systems, North-Holland, Elsevier, 1990.Google Scholar
- Thorpe S.J., Gautrais J. Rapid visual processing using spike asynchrony. NIPS’96, Denver, CO. In Jordan (Ed.), Neural Information Processing Systems, 9.Google Scholar
- VanRullen R., Delorme A., Gautrais J., Thorpe S.J. Face processing using one spike per neuron. Biosystems, in press.Google Scholar
- Samuelides M., Thorpe S.J., Veneau E. Implementing Hebbian Learning in a Rank-based Neural Network. ICANN’97, Lausanne, Switzerland, 8–10 oct. 1997.Google Scholar
- Le Cun Y., Jackel L., Bottou L., Brunot A., Cortes C., Denker J., Drucker H., Guyon I., Müller U., Säckinger E., Simard P., Vapnik V. Comparison of learning algorithms for handwritten digit recognition. ICANN’95, Paris, France, 9–13 oct. 1995.Google Scholar
- Muller J.D., Samuelides M. Integrated cooperation between unsupervised and supervised learning for a defect classification problem. ICANN’92, Brighton, UK. Artificial Neural Networks, Vol. 2, Aleksander, and Taylor (Eds.), Elsevier Science Publishers.Google Scholar
- Muller J.D., Cheynet P., Velazco R. Analysis and improvement of neural network robustness for on-board satellite image processing. ICANN’97, Lausanne, Switzerland, 8–10 oct. 1997.Google Scholar