One of the fundamental issues in machine learning is how to measure the similarity between data instances. For example, in clustering, we need to compare the distances between a data point to the cluster centres, or evaluate the densities around data points based on their distances to neighbours, the finding of which also demands distance calculation. In classification, a straightforward approach may be to search out the data entries that best match the input, and classify the input to the class that has the most votes among the searched neighbours.