Maximum Likelihood Estimation of Functionals of Discrete Distributions
We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to anal...
Ausführliche Beschreibung
Autor*in: |
Jiao, Jiantao [verfasserIn] |
---|
Format: |
Artikel |
---|---|
Sprache: |
Englisch |
Erschienen: |
2017 |
---|
Schlagwörter: |
---|
Systematik: |
|
---|
Übergeordnetes Werk: |
Enthalten in: IEEE transactions on information theory - Piscataway, NJ : IEEE, 1963, 63(2017), 10, Seite 6774-6798 |
---|---|
Übergeordnetes Werk: |
volume:63 ; year:2017 ; number:10 ; pages:6774-6798 |
Links: |
---|
DOI / URN: |
10.1109/TIT.2017.2733537 |
---|
Katalog-ID: |
OLC1996944002 |
---|
LEADER | 01000caa a2200265 4500 | ||
---|---|---|---|
001 | OLC1996944002 | ||
003 | DE-627 | ||
005 | 20220221163315.0 | ||
007 | tu | ||
008 | 171125s2017 xx ||||| 00| ||eng c | ||
024 | 7 | |a 10.1109/TIT.2017.2733537 |2 doi | |
028 | 5 | 2 | |a PQ20171228 |
035 | |a (DE-627)OLC1996944002 | ||
035 | |a (DE-599)GBVOLC1996944002 | ||
035 | |a (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 | ||
035 | |a (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
082 | 0 | 4 | |a 070 |a 620 |q DE-600 |
084 | |a SA 5570 |q AVZ |2 rvk | ||
100 | 1 | |a Jiao, Jiantao |e verfasserin |4 aut | |
245 | 1 | 0 | |a Maximum Likelihood Estimation of Functionals of Discrete Distributions |
264 | 1 | |c 2017 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ohne Hilfsmittel zu benutzen |b n |2 rdamedia | ||
338 | |a Band |b nc |2 rdacarrier | ||
520 | |a We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. | ||
650 | 4 | |a approximation theory | |
650 | 4 | |a approximation using positive linear operators | |
650 | 4 | |a high dimensional statistics | |
650 | 4 | |a maximum likelihood estimator | |
650 | 4 | |a Information theory | |
650 | 4 | |a Entropy estimation | |
650 | 4 | |a Approximation methods | |
650 | 4 | |a Maximum likelihood estimation | |
650 | 4 | |a Complexity theory | |
650 | 4 | |a Entropy | |
650 | 4 | |a Smoothing methods | |
650 | 4 | |a Rényi entropy | |
650 | 4 | |a Dirichlet prior smoothing | |
650 | 4 | |a Statistics Theory | |
650 | 4 | |a Computer Science | |
650 | 4 | |a Information Theory | |
650 | 4 | |a Mathematics | |
700 | 1 | |a Venkat, Kartik |4 oth | |
700 | 1 | |a Han, Yanjun |4 oth | |
700 | 1 | |a Weissman, Tsachy |4 oth | |
773 | 0 | 8 | |i Enthalten in |t IEEE transactions on information theory |d Piscataway, NJ : IEEE, 1963 |g 63(2017), 10, Seite 6774-6798 |w (DE-627)12954731X |w (DE-600)218505-2 |w (DE-576)01499819X |x 0018-9448 |7 nnns |
773 | 1 | 8 | |g volume:63 |g year:2017 |g number:10 |g pages:6774-6798 |
856 | 4 | 1 | |u http://dx.doi.org/10.1109/TIT.2017.2733537 |3 Volltext |
856 | 4 | 2 | |u http://ieeexplore.ieee.org/document/7997814 |
856 | 4 | 2 | |u http://arxiv.org/abs/1406.6959 |
912 | |a GBV_USEFLAG_A | ||
912 | |a SYSFLAG_A | ||
912 | |a GBV_OLC | ||
912 | |a SSG-OLC-TEC | ||
912 | |a SSG-OLC-MAT | ||
912 | |a SSG-OLC-BUB | ||
912 | |a SSG-OPC-BBI | ||
912 | |a GBV_ILN_65 | ||
912 | |a GBV_ILN_70 | ||
912 | |a GBV_ILN_2002 | ||
912 | |a GBV_ILN_2088 | ||
936 | r | v | |a SA 5570 |
951 | |a AR | ||
952 | |d 63 |j 2017 |e 10 |h 6774-6798 |
author_variant |
j j jj |
---|---|
matchkey_str |
article:00189448:2017----::aiulklhoetmtoofntoasfi |
hierarchy_sort_str |
2017 |
publishDate |
2017 |
allfields |
10.1109/TIT.2017.2733537 doi PQ20171228 (DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete DE-627 ger DE-627 rakwb eng 070 620 DE-600 SA 5570 AVZ rvk Jiao, Jiantao verfasserin aut Maximum Likelihood Estimation of Functionals of Discrete Distributions 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics Venkat, Kartik oth Han, Yanjun oth Weissman, Tsachy oth Enthalten in IEEE transactions on information theory Piscataway, NJ : IEEE, 1963 63(2017), 10, Seite 6774-6798 (DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X 0018-9448 nnns volume:63 year:2017 number:10 pages:6774-6798 http://dx.doi.org/10.1109/TIT.2017.2733537 Volltext http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 SA 5570 AR 63 2017 10 6774-6798 |
spelling |
10.1109/TIT.2017.2733537 doi PQ20171228 (DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete DE-627 ger DE-627 rakwb eng 070 620 DE-600 SA 5570 AVZ rvk Jiao, Jiantao verfasserin aut Maximum Likelihood Estimation of Functionals of Discrete Distributions 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics Venkat, Kartik oth Han, Yanjun oth Weissman, Tsachy oth Enthalten in IEEE transactions on information theory Piscataway, NJ : IEEE, 1963 63(2017), 10, Seite 6774-6798 (DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X 0018-9448 nnns volume:63 year:2017 number:10 pages:6774-6798 http://dx.doi.org/10.1109/TIT.2017.2733537 Volltext http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 SA 5570 AR 63 2017 10 6774-6798 |
allfields_unstemmed |
10.1109/TIT.2017.2733537 doi PQ20171228 (DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete DE-627 ger DE-627 rakwb eng 070 620 DE-600 SA 5570 AVZ rvk Jiao, Jiantao verfasserin aut Maximum Likelihood Estimation of Functionals of Discrete Distributions 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics Venkat, Kartik oth Han, Yanjun oth Weissman, Tsachy oth Enthalten in IEEE transactions on information theory Piscataway, NJ : IEEE, 1963 63(2017), 10, Seite 6774-6798 (DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X 0018-9448 nnns volume:63 year:2017 number:10 pages:6774-6798 http://dx.doi.org/10.1109/TIT.2017.2733537 Volltext http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 SA 5570 AR 63 2017 10 6774-6798 |
allfieldsGer |
10.1109/TIT.2017.2733537 doi PQ20171228 (DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete DE-627 ger DE-627 rakwb eng 070 620 DE-600 SA 5570 AVZ rvk Jiao, Jiantao verfasserin aut Maximum Likelihood Estimation of Functionals of Discrete Distributions 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics Venkat, Kartik oth Han, Yanjun oth Weissman, Tsachy oth Enthalten in IEEE transactions on information theory Piscataway, NJ : IEEE, 1963 63(2017), 10, Seite 6774-6798 (DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X 0018-9448 nnns volume:63 year:2017 number:10 pages:6774-6798 http://dx.doi.org/10.1109/TIT.2017.2733537 Volltext http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 SA 5570 AR 63 2017 10 6774-6798 |
allfieldsSound |
10.1109/TIT.2017.2733537 doi PQ20171228 (DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete DE-627 ger DE-627 rakwb eng 070 620 DE-600 SA 5570 AVZ rvk Jiao, Jiantao verfasserin aut Maximum Likelihood Estimation of Functionals of Discrete Distributions 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics Venkat, Kartik oth Han, Yanjun oth Weissman, Tsachy oth Enthalten in IEEE transactions on information theory Piscataway, NJ : IEEE, 1963 63(2017), 10, Seite 6774-6798 (DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X 0018-9448 nnns volume:63 year:2017 number:10 pages:6774-6798 http://dx.doi.org/10.1109/TIT.2017.2733537 Volltext http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 SA 5570 AR 63 2017 10 6774-6798 |
language |
English |
source |
Enthalten in IEEE transactions on information theory 63(2017), 10, Seite 6774-6798 volume:63 year:2017 number:10 pages:6774-6798 |
sourceStr |
Enthalten in IEEE transactions on information theory 63(2017), 10, Seite 6774-6798 volume:63 year:2017 number:10 pages:6774-6798 |
format_phy_str_mv |
Article |
institution |
findex.gbv.de |
topic_facet |
approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics |
dewey-raw |
070 |
isfreeaccess_bool |
false |
container_title |
IEEE transactions on information theory |
authorswithroles_txt_mv |
Jiao, Jiantao @@aut@@ Venkat, Kartik @@oth@@ Han, Yanjun @@oth@@ Weissman, Tsachy @@oth@@ |
publishDateDaySort_date |
2017-01-01T00:00:00Z |
hierarchy_top_id |
12954731X |
dewey-sort |
270 |
id |
OLC1996944002 |
language_de |
englisch |
fullrecord |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a2200265 4500</leader><controlfield tag="001">OLC1996944002</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20220221163315.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">171125s2017 xx ||||| 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1109/TIT.2017.2733537</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">PQ20171228</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC1996944002</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVOLC1996944002</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">070</subfield><subfield code="a">620</subfield><subfield code="q">DE-600</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">SA 5570</subfield><subfield code="q">AVZ</subfield><subfield code="2">rvk</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Jiao, Jiantao</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Maximum Likelihood Estimation of Functionals of Discrete Distributions</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2017</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">ohne Hilfsmittel zu benutzen</subfield><subfield code="b">n</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Band</subfield><subfield code="b">nc</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">approximation theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">approximation using positive linear operators</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">high dimensional statistics</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">maximum likelihood estimator</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Information theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Entropy estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Approximation methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Maximum likelihood estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Complexity theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Entropy</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Smoothing methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Rényi entropy</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Dirichlet prior smoothing</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Statistics Theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Computer Science</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Information Theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Mathematics</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Venkat, Kartik</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Han, Yanjun</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Weissman, Tsachy</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">IEEE transactions on information theory</subfield><subfield code="d">Piscataway, NJ : IEEE, 1963</subfield><subfield code="g">63(2017), 10, Seite 6774-6798</subfield><subfield code="w">(DE-627)12954731X</subfield><subfield code="w">(DE-600)218505-2</subfield><subfield code="w">(DE-576)01499819X</subfield><subfield code="x">0018-9448</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:63</subfield><subfield code="g">year:2017</subfield><subfield code="g">number:10</subfield><subfield code="g">pages:6774-6798</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">http://dx.doi.org/10.1109/TIT.2017.2733537</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://ieeexplore.ieee.org/document/7997814</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://arxiv.org/abs/1406.6959</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_OLC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-TEC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-MAT</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-BUB</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OPC-BBI</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_65</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_70</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2002</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2088</subfield></datafield><datafield tag="936" ind1="r" ind2="v"><subfield code="a">SA 5570</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">63</subfield><subfield code="j">2017</subfield><subfield code="e">10</subfield><subfield code="h">6774-6798</subfield></datafield></record></collection>
|
author |
Jiao, Jiantao |
spellingShingle |
Jiao, Jiantao ddc 070 rvk SA 5570 misc approximation theory misc approximation using positive linear operators misc high dimensional statistics misc maximum likelihood estimator misc Information theory misc Entropy estimation misc Approximation methods misc Maximum likelihood estimation misc Complexity theory misc Entropy misc Smoothing methods misc Rényi entropy misc Dirichlet prior smoothing misc Statistics Theory misc Computer Science misc Information Theory misc Mathematics Maximum Likelihood Estimation of Functionals of Discrete Distributions |
authorStr |
Jiao, Jiantao |
ppnlink_with_tag_str_mv |
@@773@@(DE-627)12954731X |
format |
Article |
dewey-ones |
070 - News media, journalism & publishing 620 - Engineering & allied operations |
delete_txt_mv |
keep |
author_role |
aut |
collection |
OLC |
remote_str |
false |
illustrated |
Not Illustrated |
issn |
0018-9448 |
topic_title |
070 620 DE-600 SA 5570 AVZ rvk Maximum Likelihood Estimation of Functionals of Discrete Distributions approximation theory approximation using positive linear operators high dimensional statistics maximum likelihood estimator Information theory Entropy estimation Approximation methods Maximum likelihood estimation Complexity theory Entropy Smoothing methods Rényi entropy Dirichlet prior smoothing Statistics Theory Computer Science Information Theory Mathematics |
topic |
ddc 070 rvk SA 5570 misc approximation theory misc approximation using positive linear operators misc high dimensional statistics misc maximum likelihood estimator misc Information theory misc Entropy estimation misc Approximation methods misc Maximum likelihood estimation misc Complexity theory misc Entropy misc Smoothing methods misc Rényi entropy misc Dirichlet prior smoothing misc Statistics Theory misc Computer Science misc Information Theory misc Mathematics |
topic_unstemmed |
ddc 070 rvk SA 5570 misc approximation theory misc approximation using positive linear operators misc high dimensional statistics misc maximum likelihood estimator misc Information theory misc Entropy estimation misc Approximation methods misc Maximum likelihood estimation misc Complexity theory misc Entropy misc Smoothing methods misc Rényi entropy misc Dirichlet prior smoothing misc Statistics Theory misc Computer Science misc Information Theory misc Mathematics |
topic_browse |
ddc 070 rvk SA 5570 misc approximation theory misc approximation using positive linear operators misc high dimensional statistics misc maximum likelihood estimator misc Information theory misc Entropy estimation misc Approximation methods misc Maximum likelihood estimation misc Complexity theory misc Entropy misc Smoothing methods misc Rényi entropy misc Dirichlet prior smoothing misc Statistics Theory misc Computer Science misc Information Theory misc Mathematics |
format_facet |
Aufsätze Gedruckte Aufsätze |
format_main_str_mv |
Text Zeitschrift/Artikel |
carriertype_str_mv |
nc |
author2_variant |
k v kv y h yh t w tw |
hierarchy_parent_title |
IEEE transactions on information theory |
hierarchy_parent_id |
12954731X |
dewey-tens |
070 - News media, journalism & publishing 620 - Engineering |
hierarchy_top_title |
IEEE transactions on information theory |
isfreeaccess_txt |
false |
familylinks_str_mv |
(DE-627)12954731X (DE-600)218505-2 (DE-576)01499819X |
title |
Maximum Likelihood Estimation of Functionals of Discrete Distributions |
ctrlnum |
(DE-627)OLC1996944002 (DE-599)GBVOLC1996944002 (PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10 (KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete |
title_full |
Maximum Likelihood Estimation of Functionals of Discrete Distributions |
author_sort |
Jiao, Jiantao |
journal |
IEEE transactions on information theory |
journalStr |
IEEE transactions on information theory |
lang_code |
eng |
isOA_bool |
false |
dewey-hundreds |
000 - Computer science, information & general works 600 - Technology |
recordtype |
marc |
publishDateSort |
2017 |
contenttype_str_mv |
txt |
container_start_page |
6774 |
author_browse |
Jiao, Jiantao |
container_volume |
63 |
class |
070 620 DE-600 SA 5570 AVZ rvk |
format_se |
Aufsätze |
author-letter |
Jiao, Jiantao |
doi_str_mv |
10.1109/TIT.2017.2733537 |
dewey-full |
070 620 |
title_sort |
maximum likelihood estimation of functionals of discrete distributions |
title_auth |
Maximum Likelihood Estimation of Functionals of Discrete Distributions |
abstract |
We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. |
abstractGer |
We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. |
abstract_unstemmed |
We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples. |
collection_details |
GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT SSG-OLC-BUB SSG-OPC-BBI GBV_ILN_65 GBV_ILN_70 GBV_ILN_2002 GBV_ILN_2088 |
container_issue |
10 |
title_short |
Maximum Likelihood Estimation of Functionals of Discrete Distributions |
url |
http://dx.doi.org/10.1109/TIT.2017.2733537 http://ieeexplore.ieee.org/document/7997814 http://arxiv.org/abs/1406.6959 |
remote_bool |
false |
author2 |
Venkat, Kartik Han, Yanjun Weissman, Tsachy |
author2Str |
Venkat, Kartik Han, Yanjun Weissman, Tsachy |
ppnlink |
12954731X |
mediatype_str_mv |
n |
isOA_txt |
false |
hochschulschrift_bool |
false |
author2_role |
oth oth oth |
doi_str |
10.1109/TIT.2017.2733537 |
up_date |
2024-07-04T01:48:21.194Z |
_version_ |
1803611224088248320 |
fullrecord_marcxml |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a2200265 4500</leader><controlfield tag="001">OLC1996944002</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20220221163315.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">171125s2017 xx ||||| 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1109/TIT.2017.2733537</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">PQ20171228</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC1996944002</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVOLC1996944002</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(PRQ)a1023-d7b684d5f7b3b3ec98e8ab70e5508eb0d03460fcbfacd4b0adbed995a54f96c10</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(KEY)0023448620170000063001006774maximumlikelihoodestimationoffunctionalsofdiscrete</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">070</subfield><subfield code="a">620</subfield><subfield code="q">DE-600</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">SA 5570</subfield><subfield code="q">AVZ</subfield><subfield code="2">rvk</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Jiao, Jiantao</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Maximum Likelihood Estimation of Functionals of Discrete Distributions</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2017</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">ohne Hilfsmittel zu benutzen</subfield><subfield code="b">n</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Band</subfield><subfield code="b">nc</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">We consider the problem of estimating functionals of discrete distributions, and focus on a tight (up to universal multiplicative constants for each specific functional) nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their bias . We explicitly characterize the worst case squared error risk incurred by the maximum likelihood estimator (MLE) in estimating the Shannon entropy <inline-formula> <tex-math notation="LaTeX">H(P) = \sum _{i = 1}^{S} -p_{i} \ln p_{i} </tex-math></inline-formula>, and the power sum <inline-formula> <tex-math notation="LaTeX">F_\alpha (P) = \sum _{i = 1}^{S} p_{i}^\alpha ,\alpha >0 </tex-math></inline-formula>, up to universal multiplicative constants for each fixed functional, for any alphabet size <inline-formula> <tex-math notation="LaTeX">S\leq \infty </tex-math></inline-formula> and sample size <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have <inline-formula> <tex-math notation="LaTeX">n \gg S </tex-math></inline-formula> observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider <inline-formula> <tex-math notation="LaTeX">n \gg S^{1/\alpha } </tex-math></inline-formula> samples for the MLE to consistently estimate <inline-formula> <tex-math notation="LaTeX">F_\alpha (P), 0<\alpha <1 </tex-math></inline-formula>. The minimax rate-optimal estimators for both problems require <inline-formula> <tex-math notation="LaTeX">S/\ln S </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">S^{1/\alpha }/\ln S </tex-math></inline-formula> samples, which implies that the MLE has a strictly sub-optimal sample complexity. When <inline-formula> <tex-math notation="LaTeX">1<\alpha <3/2 </tex-math></inline-formula>, we show that the worst case squared error rate of convergence for the MLE is <inline-formula> <tex-math notation="LaTeX">n^{-2(\alpha -1)} </tex-math></inline-formula> for infinite alphabet size, while the minimax squared error rate is <inline-formula> <tex-math notation="LaTeX">(n\ln n)^{-2(\alpha -1)} </tex-math></inline-formula>. When <inline-formula> <tex-math notation="LaTeX">\alpha \geq 3/2 </tex-math></inline-formula>, the MLE achieves the minimax optimal rate <inline-formula> <tex-math notation="LaTeX">n^{-1} </tex-math></inline-formula> regardless of the alphabet size. As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. In this context, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general such estimators do not improve over the maximum likelihood estimator. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation. The performance of the minimax rate-optimal estimator with <inline-formula> <tex-math notation="LaTeX">n </tex-math></inline-formula> samples is essentially at least as good as that of Dirichlet smoothed entropy estimators with <inline-formula> <tex-math notation="LaTeX">n\ln n </tex-math></inline-formula> samples.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">approximation theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">approximation using positive linear operators</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">high dimensional statistics</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">maximum likelihood estimator</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Information theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Entropy estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Approximation methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Maximum likelihood estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Complexity theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Entropy</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Smoothing methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Rényi entropy</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Dirichlet prior smoothing</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Statistics Theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Computer Science</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Information Theory</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Mathematics</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Venkat, Kartik</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Han, Yanjun</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Weissman, Tsachy</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">IEEE transactions on information theory</subfield><subfield code="d">Piscataway, NJ : IEEE, 1963</subfield><subfield code="g">63(2017), 10, Seite 6774-6798</subfield><subfield code="w">(DE-627)12954731X</subfield><subfield code="w">(DE-600)218505-2</subfield><subfield code="w">(DE-576)01499819X</subfield><subfield code="x">0018-9448</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:63</subfield><subfield code="g">year:2017</subfield><subfield code="g">number:10</subfield><subfield code="g">pages:6774-6798</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">http://dx.doi.org/10.1109/TIT.2017.2733537</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://ieeexplore.ieee.org/document/7997814</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://arxiv.org/abs/1406.6959</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_OLC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-TEC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-MAT</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-BUB</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OPC-BBI</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_65</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_70</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2002</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2088</subfield></datafield><datafield tag="936" ind1="r" ind2="v"><subfield code="a">SA 5570</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">63</subfield><subfield code="j">2017</subfield><subfield code="e">10</subfield><subfield code="h">6774-6798</subfield></datafield></record></collection>
|
score |
7.3996363 |