<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3244">
  <Title>Learning Nonstructural Distance Metric by Minimum Cluster Distortions</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> We proposed a global metric distance that is useful for clustering or retrieval where Euclidean distance has been used. This distance is optimal in the sense of quadratic minimization over all the clusters in the training data. Experiments on sentence retrieval, document retrieval and K-means clustering all showed improvements over Euclidean distance, with a significant refinement with tight training clusters in sentence retrieval.</Paragraph>
  </Section>
class="xml-element"></Paper>