532 research outputs found
Comment on Article by Jain and Neal
No full textArticle commenté : Splitting and Merging Components of a Nonconjugate Dirichlet Process Mixture Model, by Sonia Jain and Radford M. Neal, http://dx.doi.org/10.1214/07-BA21
MCMC Methods for Non-linear State Space Models
No full textThis thesis is concerned with developing efficient MCMC (Markov Chain Monte Carlo) techniques for non-linear, non-Gaussian state space models. These models are widely applicable in many fields, such as population ecology and economics. However, Bayesian inference in such models remains challenging, since sampling the latent variables and parameters in them can be difficult.
We begin by showing how an efficient MCMC method can be developed for a queueing model, inference for which has previously been done with ABC (Approximate Bayesian Computation). This example is an illustration of how we can do inference with MCMC in a model in which inference is considered difficult and for which ABC was used. We do this by expanding the state to include variables involved in the process used to generate the observed data.
We then proceed to the main contributions of the thesis, which are developing general MCMC methods for non-linear, non-Gaussian state space models based on the ensemble MCMC and embedded HMM MCMC frameworks. The embedded HMM and ensemble MCMC methods we introduce provide a unified framework that is sufficiently flexible to often address difficult issues in MCMC sampling for non-linear, non-Gaussian state space models, without having to resort to specialized methods for different cases.
We start with the basic embedded HMM method of Neal (2003), and Neal, Beal and Roweis (2004), designed for sampling latent state sequences given known parameter values, and develop an efficient ensemble MCMC method for sampling parameters when the parameters and the latent sequence are strongly dependent. This method is tested on a non-linear model of population dynamics. We then show how ensembles can be used to make parameter sampling more efficient in a stochastic volatility model by making use of a caching technique (Neal (2006)). Finally, we develop an ensemble MCMC method for sampling sequences in non-linear, non-Gaussian state space models when the state space is high-dimensional and test it on two sample models with high-dimensional state spaces, one of which is multimodal, and compare the performance of this method to a particle MCMC method.Ph.D
A Non-Reversible Markov Chain Sampling Method
No full textThis issue was undated. The date given is an estimate.21 pages, 1 article*A Non-Reversible Markov Chain Sampling Method* (Diaconis, Persi; Holmes, Susan; Neal, Radford) 21 page
Radford and the Splashy Fen Concert
No full textThis document discusses how excited the author is for his trip in April and the Splashy Fen concert
Takings Law and the Regulatory State: A Response to R.S. Radford
No full textIn the Winter 1994 issue of the Fordham Urban Law Journal, R.S. Radford provided an illuminating review of Dennis Coyle\u27s book Property Rights and the Constitution. Radford observes that, in addition to studying post-New Deal land use cases, Coyle provides an ideological framework that illuminates several key strands in the constitutional jurisprudence of property law ... [and] sets forth his own theories of the vital role of private property in creating and maintaining the American constitutional system. Radford\u27s review is a generally enthusiastic one. He sees Coyle\u27s book as providing a much-needed corrective to the existing pro-regulatory bias in the [scholarly] literature. He applauds Coyle, as well, for enriching our understanding of the competing preference systems that lead to different views about the legitimacy of land use regulation.
Underlying Radford\u27s review is the idea that property rights deserve greater constitutional protection than they have received in the almost sixty years since the Supreme Court accepted the fundamental legitimacy of the regulatory state. Radford\u27s position in this regard is not novel, but reflects broader trends in the courts and in the academy. In particular, Professor Richard Epstein of the University of Chicago has argued that the Fifth Amendment\u27s Takings Clause should be interpreted to bar government actions with redistributive consequences-to bar, in other words, the modern regulatory state. At the same time, in a series of recent cases involving land use and the Takings Clause, the Supreme Court has expanded the scope of the Takings Clause, although its holdings have been narrower in scope than Epstein\u27s view would warrant.
In this response, the author uses Radford\u27s review to talk about property rights and the Constitution. First, he reviews Radford\u27s interpretation and criticism of Coyle\u27s theory. Then the author discusses Radford\u27s Culture X theory in the context of Lucas v. South Carolina Coastal Council. Finally, he discusses the constitutional implications of Radford\u27s analysis
Gaussian Process Regression with Heteroscedastic Residuals and Fast MCMC Methods
No full textGaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this thesis, we propose a GP regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for the GPLC and GPLV models. Experiments show that when the data isheteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and smaller negative log-probability density) than standard GP regression. In addition, if we do not assume Gaussian residuals, our GPLC model (as in case (b) above) is still generally nearly as good as GPLV when the residuals are in fact Gaussian. When the residuals are non-Gaussian, our GPLC model is better than GPLV.Evaluating the posterior probability density function is the most costly operation when Markov Chain Monte Carlo (MCMC) is applied to many Bayesian inference problems. For GP models, computing the posterior density involves computing the covariance matrix, and then inverting the covariance matrix. The computation time for computingthe covariance matrix is proportional to pn2, and for the inversion is proportional to n3, where p is the number of covariates and n is the number of training cases. We introduce MCMC methods based on the "temporary mapping and caching" framework (Neal, 2006), using a fast approximation, *, as the distribution needed to construct thetemporary space. We propose two implementations under this scheme: "mapping to a discretizing chain", and "mapping with tempered transitions", both of which are exactly correct MCMC methods for sampling , even though their transitions are constructed using an approximation. These methods are equivalent when their tuning parameters are set at the simplest values, but differ in general. We compare how well these methods work when using several approximations, nding on synthetic datasets that a * based on the "Subset of Data" (SOD) method is almost always more efficient than standard MCMC using only On some datasets, a more sophisticated * based on the "Nystrom-Cholesky" method works better than SOD.Ph.D
Gaussian Process Regression with Heteroscedastic Residuals and Fast MCMC Methods
No full textGaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this thesis, we propose a GP regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for the GPLC and GPLV models. Experiments show that when the data isheteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and smaller negative log-probability density) than standard GP regression. In addition, if we do not assume Gaussian residuals, our GPLC model (as in case (b) above) is still generally nearly as good as GPLV when the residuals are in fact Gaussian. When the residuals are non-Gaussian, our GPLC model is better than GPLV.Evaluating the posterior probability density function is the most costly operation when Markov Chain Monte Carlo (MCMC) is applied to many Bayesian inference problems. For GP models, computing the posterior density involves computing the covariance matrix, and then inverting the covariance matrix. The computation time for computingthe covariance matrix is proportional to pn2, and for the inversion is proportional to n3, where p is the number of covariates and n is the number of training cases. We introduce MCMC methods based on the "temporary mapping and caching" framework (Neal, 2006), using a fast approximation, *, as the distribution needed to construct thetemporary space. We propose two implementations under this scheme: "mapping to a discretizing chain", and "mapping with tempered transitions", both of which are exactly correct MCMC methods for sampling , even though their transitions are constructed using an approximation. These methods are equivalent when their tuning parameters are set at the simplest values, but differ in general. We compare how well these methods work when using several approximations, nding on synthetic datasets that a * based on the "Subset of Data" (SOD) method is almost always more efficient than standard MCMC using only On some datasets, a more sophisticated * based on the "Nystrom-Cholesky" method works better than SOD.Ph.D
Reading Literary Cannibalism through Specific Body Parts
No full textKathryn Radford
Reading Literary Cannibalism through Specific Body Parts
This article outlines how the modem cannibal myth functions on the basis of prior references in Western art and literature (mythemes). By tracing the importance of the heart and brain plus the eating thereof. the author points up a semantic shift from 'sacred heart' to 'secular brain'. The cannibal reappears at the body part which represents the ultimate; in other words, ultimate act and ultimate body part, the locus of many contemporary societal preoccupations (Kuru. CJT, transplants). The article refers specifically to the trilogy of Thomas Harris. in particular, Hannibal. This is an extract of a broader study of the real act of cannibalism in twentieth-century Western literature.Kathryn Radford
Reading Literary Cannibalism through Specific Body Parts
This article outlines how the modem cannibal myth functions on the basis of prior references in Western art and literature (mythemes). By tracing the importance of the heart and brain plus the eating thereof. the author points up a semantic shift from 'sacred heart' to 'secular brain'. The cannibal reappears at the body part which represents the ultimate; in other words, ultimate act and ultimate body part, the locus of many contemporary societal preoccupations (Kuru. CJT, transplants). The article refers specifically to the trilogy of Thomas Harris. in particular, Hannibal. This is an extract of a broader study of the real act of cannibalism in twentieth-century Western literature
An Appalachian curriculum.
No full textCover title."This work, originally published by the Appalachian Consortium Press, has been reissued in an edition unaltered from its original publication. Open access editions of this and other Appalachian Consortium Press publications are available"--Back cover."An Appalachian Curriculum is the product of a week-long teachers' workshop held on the campus of Radford University in June, 1998"--Page 2.Print version record.JSTO
Correction to: Effectiveness of quality incentive payments in general practice (EQuIP-GP): a study protocol for a cluster-randomised trial of an outcomes-based funding model in Australian general practice to improve patient care (BMC health services research (2019) 19 1 (529))
No full textIn the original publication of this article [1], the first name of the 3rd author is wrong. The correct name should be Jan C. Radford. The original article has been corrected.</p
- …
