A training algorithm for optimal margin classifiers
COLT '92 Proceedings of the fifth annual workshop on Computational learning theory
Machine Learning
The recursive deterministic perceptron neural network
Neural Networks
Complexity Measures of Supervised Classification Problems
IEEE Transactions on Pattern Analysis and Machine Intelligence
Feature subset selection using a new definition of classifiability
Pattern Recognition Letters
Data complexity assessment in undersampled classification of high-dimensional biomedical data
Pattern Recognition Letters
GeoEntropy: A measure of complexity and similarity
Pattern Recognition
Current trends on knowledge extraction and neural networks
ICANN'05 Proceedings of the 15th international conference on Artificial neural networks: formal models and their applications - Volume Part II
Domain of competence of XCS classifier system in complexity measurement space
IEEE Transactions on Evolutionary Computation
The reduced nearest neighbor rule (Corresp.)
IEEE Transactions on Information Theory
IEEE Transactions on Pattern Analysis and Machine Intelligence
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
IEEE Transactions on Neural Networks
Classifiability-based omnivariate decision trees
IEEE Transactions on Neural Networks
The linear separability problem: some testing methods
IEEE Transactions on Neural Networks
Analysis of data complexity measures for classification
Expert Systems with Applications: An International Journal
On the use of meta-learning for instance selection: An architecture and an experimental study
Information Sciences: an International Journal
| Hi-index | 12.05 |
We study the relationship between linear separability and the level of complexity of classification data sets. Linearly separable classification problems are generally easier to solve than non linearly separable ones. This suggests a strong correlation between linear separability and classification complexity. We propose a novel and simple method for quantifying the complexity of the classification problem. The method, which is shown below, reduces any two class classification problem to a sequence of linearly separable steps. The number of such reduction steps could be viewed as measuring the degree of non-separability and hence the complexity of the problem. This quantification in turn can be used as a measure for the complexity of classification data sets. Results obtained using several benchmarks are provided.