About the Course:

ECE 236A is a course which covering linear optimization. Specific topics include geometry of linear programming, duality, simplex method, interior-point methods, decomposition and large-scale linear programming, quadratic programming and complementary pivot theory.

Abstract:

This project examines the significance of the sample size used when training a support vector machine (SVM) classifier. A classifier is trained using vectors with associated labels. The objective is to predict the label of a vector given the entries of that vector. The accuracy of the classifier is determined by averaging the amount of correctly predicted labels with respect to the total amount. Here, we consider a classifier with a central node that receives constrained communication from distributed sensors and that will train a classifier using data points obtained from those sensors. Yet, there is a cost, that the classifier will be attempting to minimize, associated with each data point transmission. Some data points will prove to be more “useful” than others when training the classifier.