Simple approximate MAP inference for Dirichlet processes mixtures
- Submitting institution
-
The University of Birmingham
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 77471835
- Type
- D - Journal article
- DOI
-
10.1214/16-EJS1196
- Title of journal
- Electronic Journal of Statistics
- Article number
- -
- First page
- 3548
- Volume
- 10
- Issue
- 2
- ISSN
- 1935-7524
- Open access status
- Compliant
- Month of publication
- November
- Year of publication
- 2016
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
2
- Research group(s)
-
-
- Citation count
- 10
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Bayesian nonparametrics, including Dirichlet process modelling, is an important field of analytical techniques used in data science, statistics, and machine learning. However, these techniques require heavy computational resources and are therefore generally impractical for modern, large-scale data analysis problems. This paper provides an important theoretical advance: replacing 'expensive' Markov chain Monte-Carlo inference with 'cheap' iterative conditional modes and appropriate maximum likelihood procedures, making Bayesian nonparametrics feasible for modern, large-scale data problems. Published in an internationally-leading theoretical statistics journal, this work has already enabled novel applications of Bayesian nonparametrics across multiple disciplines, including artificial intelligence, statistics, sensor data processing, and bioinformatics.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
