On measure concentration of random maximum a-posteriori perturbations
Name
Jaakkola_On measure.pdf
Size
344.88 KB
Format
Adobe PDF
Checksum (MD5)
01345be3cb82c2267168d89e5b2848e8
Author(s)
Orabona, Francesco
Hazan, Tamir
Sarwate, Anand D.
Jaakkola, Tommi S.
Date Issued
2014
Journal
Journal of Machine Learning Research
Publisher
Association for Computing Machinery (ACM)
Citation
Orabona, Francesco, Tamir Hazan, Anand D. Sarwate, and Tommi S. Jaakkola. "On measure concentration of random maximum a-posteriori perturbations." Journal of Machine Learning Research, Volume 32: Proceedings of The 31st International Conference on Machine Learning (2014).
Version
Final published version
Abstract
The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for inference and learning in high dimensional complex models. By maximizing a randomly perturbed potential function, MAP perturbations generate unbiased samples from the Gibbs distribution. Unfortunately, the computational cost of generating so many high-dimensional random variables can be prohibitive. More efficient algorithms use sequential sampling strategies based on the expected value of low dimensional MAP perturbations. This paper develops new measure concentration inequalities that bound the number of samples needed to estimate such expected values. Applying the general result to MAP perturbations can yield a more efficient algorithm to approximate sampling from the Gibbs distribution. The measure concentration result is of general interest and may be applicable to other areas involving Monte Carlo estimation of expectations.
MIT Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
http://jmlr.org/proceedings/papers/v32/
