This paper studies on the mirror image learning algorithm for the autoassociative neural networks and evaluates the performance by handwritten numeral recognition test. Each of th...
A class-modular generalized learning vector quantization (GLVQ) ensemble method with outlier learning for handwritten digit recognition is proposed. A GLVQ classifier is one of d...
In style-constrained classification often there are only a few samples of each style and class, and the correspondences between styles in the training set and the test set are un...
This paper describes our investigation into the neural gas (NG) network algorithm and the hierarchical overlapped architecture (HONG) which we have built by retaining the essence ...
84.6% of touching digit strings have only two digits touching, 12.3% have three digits touching, and 3.1% have more than three digits touching. We present a multiexperts approach ...