Condensed Nearest Neighbor Method

We find a few representative samples to reduce storage.

1. Separate ${\mathcal{X}}^N$ as ${\mathcal{X}}_S$ and ${\mathcal{X}}_G$.
2. Start with only 1 sample in ${\mathcal{X}}_S$ and all others are in ${\mathcal{X}}_G$.
3. Iterate through each sample in ${\mathcal{X}}_G$, if it can be correctly classified with ${\mathcal{X}}_S$ then stay. If not, move to ${\mathcal{X}}_G$.
4. Use only samples in ${\mathcal{X}}_S$ to classify the test set.