Generalization of shift invariant neural networks: Image processing of corneal endothelium

Abstract In training neural networks, if a training set is not randomly extracted from the whole input domain, it is difficult to construct well generalized neural networks. In this paper, to solve the problem in image processing of corneal endothelium, we propose a method to get a well generalized shift-invariant neural network by designing optimal weights between the input and first hidden layers. The neural network has weight patterns between each layer, and its signal propagation is calculated by discrete convolution between the weight patterns and data in lower layers. The optimal weights between the input and first hidden layer are designed by observing the training set distribution in the vector space. The designed weights transform the training set to cover the whole input domain. In the learning process, the designed weights are fixed and the other weights are updated. In the experiment, well generalized neural networks are obtained.

[1]  Michael C. Mozer,et al.  Using Relevance to Reduce Network Size Automatically , 1989 .

[2]  Kazuyoshi Itoh,et al.  Detection of Cell Membranes in Human Corneal Endothelial Micrograms Using Mathematical Morphology , 1992 .

[3]  Osamu Matoba,et al.  Learning generalization by validation set , 1992 .

[4]  K Doi,et al.  Computerized detection of clustered microcalcifications in digital mammograms using a shift-invariant artificial neural network. , 1994, Medical physics.

[5]  Jack Sklansky,et al.  Mapping Techniques for Exploratory Pattern Analysis , 1988 .

[6]  Lawrence D. Jackel,et al.  Backpropagation Applied to Handwritten Zip Code Recognition , 1989, Neural Computation.

[7]  Matthew T. Freedman,et al.  Convolution neural-network-based detection of lung structures , 1994, Medical Imaging.

[8]  Emil Wolf,et al.  Principles of Optics: Contents , 1999 .

[9]  David S. Touretzky,et al.  Connectionist models : proceedings of the 1990 summer school , 1991 .

[10]  Yann LeCun,et al.  Optimal Brain Damage , 1989, NIPS.

[11]  Y Ichioka,et al.  Parallel distributed processing model with local space-invariant interconnections and its optical architecture. , 1990, Applied optics.

[12]  Wei Zhang,et al.  Image processing of human corneal endothelium based on a learning network. , 1991, Applied optics.

[13]  Chuanyi Ji,et al.  Generalizing Smoothness Constraints from Discrete Samples , 1990, Neural Computation.

[14]  Erkki Oja,et al.  Subspace methods of pattern recognition , 1983 .

[15]  David E. Rumelhart,et al.  BACK-PROPAGATION, WEIGHT-ELIMINATION AND TIME SERIES PREDICTION , 1991 .

[16]  Joseph J. Atick,et al.  What Does the Retina Know about Natural Scenes? , 1992, Neural Computation.