A rate-distortion one-class model and its applications to clustering

In one-class classification we seek a rule to find a coherent subset of instances similar to a few positive examples in a large pool of instances. The problem can be formulated and analyzed naturally in a rate-distortion framework, leading to an efficient algorithm that compares well with two previous oneclass methods. The model can be also be extended to remove background clutter in clustering to improve cluster purity.

• Type: Conference
• Year: 2008
• Where: ICML
• Authors: Koby Crammer, Partha Pratim Talukdar, Fernando Pereira
• Keywords: Cluster Purity, Coherent Subset, ICML 2008, Machine Learning, Rate-distortion Framework