Some question to Monte-Carlo simulation in AIB algorithm

Authors:
Sanming Song;Qunsheng Yang;Yinwei Zhan
Affiliations:
Faculty of Computer, Guangdong University of Technology, Guangzhou, P. R. China;Faculty of Computer, Guangdong University of Technology, Guangzhou, P. R. China;Faculty of Computer, Guangdong University of Technology, Guangzhou, P. R. China
Venue:
AIRS'08 Proceedings of the 4th Asia information retrieval conference on Information retrieval technology
Year:
2008

Citing 6
Cited 0

Learning mixture hierarchies

Proceedings of the 1998 conference on Advances in neural information processing systems II
Unsupervised Image Clustering Using the Information Bottleneck Method

Proceedings of the 24th DAGM Symposium on Pattern Recognition
An Efficient Image Similarity Measure Based on Approximations of KL-Divergence Between Two Gaussian Mixtures

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Applying the Information Bottleneck Principle to Unsupervised Clustering of Discrete and Continuous Image Representations

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
On the efficient evaluation of probabilistic similarity functions for image retrieval

IEEE Transactions on Information Theory
Unsupervised image-set clustering using an information theoretic framework

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Hierarchical clustering algorithm is efficient in reducing the bytes needed to describe the original information while preserving the original information structure. Information Bottleneck (IB) theory is a hierarchical clustering framework derivative from the information theory. Agglomerative Information Bottleneck (AIB) algorithm is a suboptimal agglomerative clustering procedure designed for optimizing the original computation-exhausted IB algorithm. But the Monte-Carlo simulation formula which is widely adopted for distortion measures in AIB algorithm is problematic. This paper testified that there being a contradiction between the adopted Monte-Carlo formula and IB principle. Extending special distortion measures to common distances, the paper also present several proposals. And Experiments show their efficiency and availability.