Machine learning, neural and statistical classification
Machine learning, neural and statistical classification
Neural Networks for Pattern Recognition
Neural Networks for Pattern Recognition
Evolution strategies –A comprehensive introduction
Natural Computing: an international journal
In Defense of One-Vs-All Classification
The Journal of Machine Learning Research
A Projection Pursuit Algorithm for Exploratory Data Analysis
IEEE Transactions on Computers
Solving multiclass learning problems via error-correcting output codes
Journal of Artificial Intelligence Research
Hi-index | 0.00 |
Neural networks for classification require our choosing output codes. Most often, the 1-of-c output code is used, with as many dimensions as classes. This coding scheme can turn into a burden for datasets with many classes such as the 19 class UCI soybean problem. In this paper, a procedure is introduced which allows to choose the number of output units of a neural network, independently of the number of classes. The weights of the network are learned by means of an evolution strategy whose fitness is the number of misclassifications incurred by assigning patterns to the class of the closest projected mean. In this way, we obtain two-dimensional views of multiclass problems such as the 19-class soybean database and the non-linear 6-class satellite problem.