Statistical Translation, Heat Kernels and Expected Distances
Machine Learning
2012-06-26 v1 Computer Vision and Pattern Recognition
Information Retrieval
Machine Learning
Abstract
High dimensional structured data such as text and images is often poorly understood and misrepresented in statistical modeling. The standard histogram representation suffers from high variance and performs poorly in general. We explore novel connections between statistical translation, heat kernels on manifolds and graphs, and expected distances. These connections provide a new framework for unsupervised metric learning for text documents. Experiments indicate that the resulting distances are generally superior to their more standard counterparts.
Cite
@article{arxiv.1206.5248,
title = {Statistical Translation, Heat Kernels and Expected Distances},
author = {Joshua Dillon and Yi Mao and Guy Lebanon and Jian Zhang},
journal= {arXiv preprint arXiv:1206.5248},
year = {2012}
}
Comments
Appears in Proceedings of the Twenty-Third Conference on Uncertainty in Artificial Intelligence (UAI2007)