Neocognitron with improved bend-extractors

We (1988) have reported previously that the performance of a neocognitron can be improved by a built-in bend-extracting layer. The conventional bend-extracting layer can detect bend points and end points of lines correctly, but not always crossing points of lines. This paper discusses that an introduction of a mechanism of disinhibition can make the bend-extracting layer detect not only bend points and end points but also crossing points of lines correctly. A neocognitron with this improved bend-extracting layer can recognize handwritten digits in the real world with a recognition rate of 98%.

[1]  Kunihiko Fukushima,et al.  Improved neocognitron with bend-detecting cells , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[2]  K Fukushima,et al.  Handwritten alphanumeric character recognition by the neocognitron , 1991, IEEE Trans. Neural Networks.

[3]  Masashi Tanigawa,et al.  Use of different thresholds in learning and recognition , 1996, Neurocomputing.

[4]  Kunihiko Fukushima,et al.  Neocognitron: A hierarchical neural network capable of visual pattern recognition , 1988, Neural Networks.

[5]  Hayaru Shouno,et al.  Training neocognitron to recognize handwritten digits in the real world , 1997, Proceedings of IEEE International Symposium on Parallel Algorithms Architecture Synthesis.

[6]  D H HUBEL,et al.  RECEPTIVE FIELDS AND FUNCTIONAL ARCHITECTURE IN TWO NONSTRIATE VISUAL AREAS (18 AND 19) OF THE CAT. , 1965, Journal of neurophysiology.