Dimensionality reduction using a novel neural network based feature extraction method

A neural network based method for feature extraction is proposed. The method achieves dimensionality reduction of input vectors used for supervised learning problems. Combinations of the original features are formed that maximize the sensitivity of the network's outputs with respect to variations of its inputs. The method exhibits some similarity to principal component analysis, but also takes into account the supervised character of the learning task. It is applied to classification problems leading to efficient dimensionality reduction and increased generalization ability.