@conference {40, title = {Multinomial Squared Direction Cosines Regression}, booktitle = {Neural Networks (IJCNN), The 2011 International Joint Conference on}, year = {2011}, note = {

Acceptance rate 75\% (468/620).

}, month = {July}, publisher = {Institute of Electrical and Electronics Engineers (IEEE)}, organization = {Institute of Electrical and Electronics Engineers (IEEE)}, address = {San Jose, California, USA }, abstract = {

In this paper we introduce Multinomial Squared Direction Cosines Regression as an alternative Multinomial Response Model. The proposed model offers an intuitive geometric interpretation to the task of estimating posterior class probabilities in multi-class problems. In specific, the latter probabilities correspond to the squared direction cosines between a given pattern and representative class exemplars that form a basis in the decision space. We demonstrate that the model allows for efficient training via a trust region based Newton{\textquoteright}s Method, provided that the number of model parameters is not too large. Experimental results on several benchmark classification problems show that it compares competitively against Logistic Regression Classifiers, Support Vector Machines, and Classification and Regression Tree models.

}, keywords = {benchmark classification problem, Computational modeling, decision space, decision theory, geometric interpretation, Kernel, logistic regression classifier, Mathematical model, model parameter, multiclass problem, multinomial response model, multinomial squared direction cosines regression, Newton method, pattern classification, regression analysis, Regression tree analysis, regression tree model, support vector machine, Support vector machines, Training}, doi = {10.1109/IJCNN.2011.6033620}, author = {Naveed H. Iqbal and Georgios C. Anagnostopoulos} }