Generalised information criteria in model selection

Konishi, Satoshi; Kitagawa, Genshiro

doi:10.1093/biomet/83.4.875

Cited by 392 publications

(323 citation statements)

References 29 publications

Supporting

Mentioning

321

Contrasting

Unclassified

Order By: Relevance

“…The parameter learning process should be therefore repeated a number of times to obtain a more stable solution. This process is also known as bootstrapping in statistics (Efron, 1979;Konishi and Kitagawa, 1996). The medians of the parameters were then calculated across repetitions for each functional category produced by each speaker.…”

Section: Analysis-by-synthesis With Stochastic Optimizationmentioning

confidence: 99%

Toward invariant functional representations of variable surface fundamental frequency contours: Synthesizing speech melody via model-based stochastic learning

Prom-on

2014

Speech Communication

View full text Add to dashboard Cite

Variability has been one of the major challenges for both theoretical understanding and computer synthesis of speech prosody. In this paper we show that economical representation of variability is the key to effective modeling of prosody. Specifically, we report the development of PENTAtrainer -A trainable yet deterministic prosody synthesizer based on an articulatory-functional view of speech. We show with testing results on Thai, Mandarin and English that it is possible to achieve high-accuracy predictive synthesis of fundamental frequency contours with very small sets of parameters obtained through stochastic learning from real speech data. The first key component of this system is syllable-synchronized sequential target approximation -implemented as the qTA model, which is designed to simulate, for each tonal unit, a wide range of contextual variability with a single invariant target. The second key component is the automatic learning of function-specific targets through stochastic global optimization, guided by a layered pseudo-hierarchical functional annotation scheme, which requires the manual labeling of only the temporal domains of the functional units. The results in terms of synthesis accuracy demonstrate that effective modeling of the contextual variability is the key also to effective modeling of function-related variability. Additionally, we show that, being both theory-based and trainable (hence datadriven), computational systems like PENTAtrainer can serve as an effective modeling tool in basic research, with which the level of falsifiability in theory testing can be raised, and also a closer link between basic and applied research in speech science can be developed.3 Graphical Abstract Highlights (maximum 85 characters/bullet)• High synthetic accuracy of prosody achieved for Thai, Mandarin and English • Many-to-one mapping from contextually variable surface F 0 to invariant functional targets • Effectively handling of both contextual and non-contextual variability • Combination of deterministic synthesis and data-driven parameter learning • Large-scale and full-detailed prosody synthesis as tool for theory testing • Freely available as a Praat scripts and plug-ins to the speech science community at large

show abstract

Section: Analysis-by-synthesis With Stochastic Optimizationmentioning

confidence: 99%

Toward invariant functional representations of variable surface fundamental frequency contours: Synthesizing speech melody via model-based stochastic learning

Prom-on

2014

Speech Communication

View full text Add to dashboard Cite

show abstract

“…The subject was first proposed by Akaike [1973], who introduced the principle of maximum entropy as the theoretical basis for model selection, and by Schwarz [1978], who, by developing a similar idea in a Bayesian context, proposed the Bayesian information criterion for model selection. Extensions of these methods include corrections to be used with small sample size [Hurvich and Tsai, 1989] and generalizations of the mentioned criteria [e.g., see Bozdogan, 1987;Konishi and Kitagawa, 1996;Wasserman, 2000]. Only with the recent increase of computer capabilities have other methods been proposed and developed for the nonasymptotic model selection, based on bootstrap [Chung et al, 1996] or on cross-validation [Browne, 2000] techniques.…”

Section: Introductionmentioning

confidence: 99%

Model selection techniques for the frequency analysis of hydrological extremes

Laio

Baldassarre

Montanari

2009

Water Resources Research

170

116

View full text Add to dashboard Cite

[1] The frequency analysis of hydrological extremes requires fitting a probability distribution to the observed data to suitably represent the frequency of occurrence of rare events. The choice of the model to be used for statistical inference is often based on subjective criteria, or it is considered a matter of probabilistic hypotheses testing. In contrast, specific tools for model selection, like the well-known Akaike information criterion (AIC) and the Bayesian information criterion (BIC), are seldom used in hydrological applications. The objective of this study is to verify whether the AIC and BIC work correctly when they are applied to identifying the probability distribution of hydrological extremes, i.e., when the available samples are small and the parent distribution is highly asymmetric. An additional model selection criterion, based on the Anderson-Darling goodness-of-fit test statistic, is here proposed, and the performances of the three methods are compared through an extensive numerical analysis. The capability of the three criteria to recognize the correct parent distribution from the available data samples varies from case to case, and it is rather good in some cases (in particular when the parent is a two-parameter distribution) and unsatisfactory in others. An application to flood peak time series from 1000 catchments located in the United Kingdom provides some further information on the qualities and drawbacks of the considered criteria. From the numerical simulations and data-based analyses it can be concluded that the three model selection techniques considered here produce results of comparable quality.

show abstract

“…In other settings, however, these criteria's tendency to err toward non-detection of true predictors may represent a serious limitation. Another concern is the variability of our methods' overoptimism estimates, which cannot be reduced by the technique of Konishi and Kitagawa (1996) since that device is applicable only in the fixed-model case. In ongoing research, we are seeking ways to surmount these limitations.…”

Section: Discussionmentioning

confidence: 99%

“…(See also Konishi and Kitagawa (1996), and the bootstrap model selection criterion of Shao (1996), which is based on prediction error loss rather than likelihood.) Suppose we sample n pairs (y i , x i ) from the data, with replacement, B times, and denote the bth bootstrap data set thus generated by (y * b , X * b ) and the associated MLEs byβ…”

Section: Selection Bias In Best-subset Regressionmentioning

confidence: 99%

Resampling-based information criteria for best-subset regression

et al. 2012

View full text Add to dashboard Cite

When a linear model is chosen by searching for the best subset among a set of candidate predictors, a fixed penalty such as that imposed by the Akaike information criterion may penalize model complexity inadequately, leading to biased model selection. We study resampling-based information criteria that aim to overcome this problem through improved estimation of the effective model dimension. The first proposed approach builds upon previous work on bootstrap-based model selection. We then propose a more novel approach based on cross-validation. Simulations and analyses of a functional neuroimaging data set illustrate the strong performance of our resamplingbased methods, which are implemented in a new R package.

show abstract

Generalised information criteria in model selection

Cited by 392 publications

References 29 publications

Toward invariant functional representations of variable surface fundamental frequency contours: Synthesizing speech melody via model-based stochastic learning

Toward invariant functional representations of variable surface fundamental frequency contours: Synthesizing speech melody via model-based stochastic learning

Model selection techniques for the frequency analysis of hydrological extremes

Resampling-based information criteria for best-subset regression

Contact Info

Product

Resources

About