Robust Neural Network Classifier
- Ali Refaee Abd Ellah Mohamed
- Mohamed Hassan Essai
- Mohamed Mohamed el- Sayed Zahra
Robust Statistics, Feed-Forward Neural Networks, M-Estimators, Classification, Robust classifier.
Classification is a data mining technique used to predict Patterns’ membership. Pattern classification involves building a
function that maps the input feature space to an output space of two or more than two classes. Neural Networks (NN) are an effective
tool in the field of pattern classification. The success of NN is highly dependent on the performance of the training process and hence
the training algorithm. Many training algorithms have been proposed so far to improve the performance of neural networks. Usually a
traditional backpropagation learning algorithm (BPLA), which minimizes the mean squared error (MSE – cost function) of the training
data, be used in the process of training neural networks. However (MSE) based learning algorithm is not robust in presence of outliers
that may pollute the training data. In our work we aim to present another cost functions which backpropagation learning algorithm
based on in order to improve the robustness of neural network training by employing a family of robust statistics estimators, commonly
known as M-estimators, and hence obtain robust NN classifiers. Comparative study between robust classifiers and non-robust
(traditional) classifiers was established in paper using crab classification problem.
Ali Refaee Abd Ellah Mohamed, Mohamed Hassan Essai, Mohamed Mohamed el- Sayed Zahra. "Robust Neural Network Classifier".INTERNATIONAL JOURNAL OF ENGINEERING DEVELOPMENT AND RESEARCH ISSN:2321-9939, Vol.1, Issue 3, pp.0, URL :https://rjwave.org/ijedr/papers/IJEDR1303065.pdf
Volume 1 Issue 3
Pages. 0