I was going through this paper which proposes a new similarity metric. The evaluation is carried out using various classification and clustering techniques. I was confused about how a similarity metric is used in classification and clustering. The similarity metric may be used as a kernel in SVM along with the linear and rbf kernel. How is the metric used with naive bayes?. Is the metric used as a distance function in clustering techniques?.
As I am new to machine learning, I am confused. Can somebody clarify how are classifiers and clustering algorithms used to evaluate a similarity metric?