Search

Your search keyword '"62B10 94A20 94A15 62E17"' showing total 2 results
Start Over Descriptor "62B10 94A20 94A15 62E17"
2 results on '"62B10 94A20 94A15 62E17"'

1. Finite-sample risk bounds for maximum likelihood estimation with arbitrary penalties

Author

Brinda, W. D. and Klusowski, Jason M.

Subjects
Mathematics - Statistics Theory, Statistics - Machine Learning, 62B10 94A20 94A15 62E17
Abstract

The MDL two-part coding $ \textit{index of resolvability} $ provides a finite-sample upper bound on the statistical risk of penalized likelihood estimators over countable models. However, the bound does not apply to unpenalized maximum likelihood estimation or procedures with exceedingly small penalties. In this paper, we point out a more general inequality that holds for arbitrary penalties. In addition, this approach makes it possible to derive exact risk bounds of order $1/n$ for iid parametric models, which improves on the order $(\log n)/n$ resolvability bounds. We conclude by discussing implications for adaptive estimation., Comment: To appear in IEEE Transactions on Information Theory, 2018

Published
2017

2. Finite-Sample Risk Bounds for Maximum Likelihood Estimation With Arbitrary Penalties

Author

W. D. Brinda and Jason M. Klusowski

Subjects
FOS: Computer and information sciences, 62B10 94A20 94A15 62E17, Machine Learning (stat.ML), Mathematics - Statistics Theory, Statistics Theory (math.ST), 02 engineering and technology, Library and Information Sciences, 01 natural sciences, Upper and lower bounds, Data modeling, 010104 statistics & probability, Statistics - Machine Learning, FOS: Mathematics, 0202 electrical engineering, electronic engineering, information engineering, Applied mathematics, Countable set, Entropy (information theory), 0101 mathematics, Minimum description length, Mathematics, Estimator, 020206 networking & telecommunications, Binary logarithm, Computer Science Applications, Parametric model, Information Systems
Abstract

The MDL two-part coding $ \textit{index of resolvability} $ provides a finite-sample upper bound on the statistical risk of penalized likelihood estimators over countable models. However, the bound does not apply to unpenalized maximum likelihood estimation or procedures with exceedingly small penalties. In this paper, we point out a more general inequality that holds for arbitrary penalties. In addition, this approach makes it possible to derive exact risk bounds of order $1/n$ for iid parametric models, which improves on the order $(\log n)/n$ resolvability bounds. We conclude by discussing implications for adaptive estimation., To appear in IEEE Transactions on Information Theory, 2018

Published
2018

Catalog

Books, media, physical & digital resources
See catalog results