Exploiting Cluster Structure to Predict The Labeling of a Graph

The nearest neighbor and the perceptron algorithms are intuitively motivated by the aims to exploit the "cluster" and "linear separation" structure of the data to be classified, respectively. We devel