• The Korean Journal of Applied Statistics
• /
• v.34 no.1
• /
• pp.9-23
• /
• 2021

In the following paper we introduce a variational Bayes method that approximates posterior distributions with mean-field method. In particular, we introduce automatic differentiation variation inference (ADVI), which approximates joint posterior distributions using the product of Gaussian distributions after transforming parameters into real coordinate space, and then apply it to pharmacokinetic models that are models for the study of the time course of drug absorption, distribution, metabolism and excretion. We analyze real data sets using ADVI and compare the results with those based on Markov chain Monte Carlo. We implement the algorithms using Stan.

• Radiation Oncology Journal
• /
• v.1 no.1
• /
• pp.25-28
• /
• 1983

Anterior and posterior parallel opposed field technique covering entire neck is desirable for elimination of junctional problems in treating upper neck with bilateral parallel opposed fields and lower neck with anterior one field. For good dose distribution in neck, dose in build-up region should be high for anterior field and should be low for posterior field. And so, with 10 MV X-ray, lead absorber was used for anterior field only. The adequate thickness of lead absorber, absorber-skin separation, width of central cord block for posterior field and anterior & posterior field weight were studied using film dosimery. The results are as follows. 1. As the thickness of the lead absorber increased the dose in build-up region increased. 2. As the absorber-skin separation decreased the dose in build-up region increased. 3. The adequate thickness of lead absorber was around 5.6mm. 4. The adequate absorber-skin separation was around 5cm. 5. The adequate posterior cord block width was 3cm. 6. 4:1 weighting for anterior and posterior field was adequate. And so with this technique, adequate dose distribution could be made as well as elimination junction problems.

• Restorative Dentistry and Endodontics
• /
• v.12 no.2
• /
• pp.33-44
• /
• 1987

The use of composite resin for the posterior teeth gives rise to clinical problem due to the lack of mechanical properties. The purpose of this study was to observe the fractured surfaces of light posterior composite resins which are P-10, Clearfil posterior, Adaptic anterior & posterior, P-30, Lite-fil posterior, Estilux posterior, Helio-molar, and Ful-fil com pules (Table 1). The failure of composite resin specimens of I, T and Y-Type (Fig. 1,2) occured under compression. Fractographical observations by SEM (JSM-T20, JEOL) were carried out in order to examine the fracture behaviour of eight composite resins in different types of specimens. The results obtained from this study were as follows: 1. Similar features were found in fractured surfaces of eight composite resins. 2. The crack growth was initiated at the regions of porosities. 3. The crack propagated on the filler-matrix interface. 4. As the crack increased in size, it accelerated to form secondary crack. 5. The fracture behaviour was dependent on the content, size, shape, and distribution of fillers.

• The Journal of Korean Academy of Prosthodontics
• /
• v.32 no.2
• /
• pp.327-353
• /
• 1994

This study was performed to evaluate the effects of number and alignment of implant fixture and various bar designs on the retention of denture and the stress distribution. Six kinds of photoelastic mandibular models and nine kinds of overdenture specimens were designed. A unilateral vertical load was gradually applied on the right first molar to calculate the maximal dislodgement load of each specimen. A unilateral vertical load of 17 Kgf was applied on the right first molar and a vertical load of 10 Kgf was applied on the interincisal edge region. The stress pattern which developed in each photoelastic model was analyzed by the reflection polariscope. The results obtained were as follows: 1. The maximal dislodgement load reversely increased with the distance from the loading point to the implant fixture, while it linearly increased with that from the most posterior implant fixture to the mesial clip. The maximal dislodgement load also increased with the use of a cantilever bar. 2. Under the posterior vertical load, the stress to the supporting tissue of the denture base increased with the distance from the loading point to the implant future. The stress concentration on the apical area of the implant future reversely increased with the distance from the loading point to the implant future. 3. In the overdentures supported by two implant fixtures under the posterior vertical load. the specimen implanted on lateral incisor areas with a cantilever bar exhibited more favorable stress distribution than that without a cantilever bar. The specimen implanted on the canine areas without a cantilever bar, however, exhibited more favorable stress distribution. 4. In the overdentures supported by three implant fixtures. the specimen implanted ell the midline and canine areas exhibited more favorable stress distribution than that implanted oil the midline and the first premolar areas. 5. In the overdentures supported by four implant fixtures. the specimen implanted with two adjacent implant fixtures exhibited more favorable stress distribution than that implanted at equal distance under the posterior vertical load. 6. Under the anterior vertical load, the overdentures supported by three implant fixtures exhibited stress concentration on the supporting structure of the middle implant future. In overdentures supported by two or four implant futures, no significant difference was noted in stress distribution between the types of bars. These results indicate that the greater the number of implant fixtures, the better the stress distribution is. A favorable stress distribution may be obtained in the overdentures supported by two or three implant fixtures, if the location and the design of the bar are appropriate.

• Phonetics and Speech Sciences
• /
• v.6 no.3
• /
• pp.155-164
• /
• 2014

This paper proposes a new method to determine the recognition units for large vocabulary continuous speech recognition (LVCSR) in Korean by applying unsupervised segmentation and merging. In the proposed method, a text sentence is segmented into morphemes and position information is added to morphemes. Then submorpheme units are obtained by splitting the morpheme units through the maximization of posterior probability terms. The posterior probability terms are computed from the morpheme frequency distribution, the morpheme length distribution, and the morpheme frequency-of-frequency distribution. Finally, the recognition units are obtained by sequentially merging the submorpheme pair with the highest frequency. Computer experiments are conducted using a Korean LVCSR with a 100k word vocabulary and a trigram language model obtained by a 300 million eojeol (word phrase) corpus. The proposed method is shown to reduce the out-of-vocabulary rate to 1.8% and reduce the syllable error rate relatively by 14.0%.

• Journal of the Korean Statistical Society
• /
• v.28 no.1
• /
• pp.93-106
• /
• 1999

In this paper we consider estimation of a real valued parameter in the drift coefficient of a Hilbert space valued Ito stochastic differential equation. First we consider observation of the corresponding diffusion in a fixed time interval [0, T] and prove the Bernstein - von Mises theorem concerning the convergence of posterior distribution of the parameter given the observation, suitably normalised and centered at the MLE, to the normal distribution as Tlongrightarrow$\infty$. As a consequence, the Bayes estimator of the drift parameter becomes asymptotically efficient and asymptotically equivalent to the MLE as Tlongrightarrow$\infty$. Next, we consider observation in a random time interval where the random time is determined by a predetermined level of precision. We show that the sequential MLE is better than the ordinary MLE in the sense that the former is unbiased, uniformly normally distributed and efficient but is latter is not so.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.20 no.3
• /
• pp.105-122
• /
• 1995

Based on linear models, the inference about the true measurement x$_{f}$ and the optimal designs x (nx1) for the calibration experiments are considered via Baysian statistical decision analysis. The posterior distribution of x$_{f}$ given the observation y$_{f}$ (qxl) and the calibration experiment is obtained with normal priors for x$_{f}$ and for themodel parameters (.alpha., .betha.). This posterior distribution is not in the form of any known distributions, which leads to the use of a numerical integration or an approximation for the calculation of the overall expected loss. The general structure of the expected loss function is characterized in the form of a conjecture. A near-optimal design is obtained through the approximation nof the conditional covariance matrix of the joint distribution of (x$_{f}$ , y$_{f}$ $^{T}$ )$^{T}$ . Numerical results for the univariate case are given to demonstrate the conjecture and to evaluate the approximation.n.

• The Journal of the Korea Contents Association
• /
• v.21 no.11
• /
• pp.77-86
• /
• 2021

A method of constructing a war simulation based on Bayesian Inference was proposed as a method of constructing heterogeneous historical war data obtained with a time difference into a single model. A method of applying a linear regression model can be considered as a method of predicting future battles by analyzing historical war results. However it is not appropriate for two heterogeneous types of historical data that reflect changes in the battlefield environment due to different times to be suitable as a single linear regression model and violation of the model's assumptions. To resolve these problems a Bayesian inference method was proposed to obtain a post-distribution by assuming the data from the previous era as a non-informative prior distribution and to infer the final posterior distribution by using it as a prior distribution to analyze the data obtained from the next era. Another advantage of the Bayesian inference method is that the results sampled by the Markov Chain Monte Carlo method can be used to infer posterior distribution or posterior predictive distribution reflecting uncertainty. In this way, it has the advantage of not only being able to utilize a variety of information rather than analyzing it with a classical linear regression model, but also continuing to update the model by reflecting additional data obtained in the future.

• Journal of the Korean Statistical Society
• /
• v.36 no.2
• /
• pp.321-333
• /
• 2007

In the proportional hazard model with the beta process prior, the posterior computation with the discrete approximation is considered. The time period of interest is partitioned by small intervals. On each partitioning interval, the likelihood is approximated by that of a binomial experiment and the beta process prior is by a beta distribution. Consequently, the posterior is approximated by that of many independent binomial model with beta priors. The analysis of the leukemia remission data is given as an example. It is illustrated that the length of the partitioning interval affects the posterior and one needs to be careful in choosing it.

• Communications for Statistical Applications and Methods
• /
• v.21 no.3
• /
• pp.261-274
• /
• 2014

In this paper we develop Bayes and empirical Bayes estimators of the finite population mean with the assumption of posterior linearity rather than normality of the superpopulation under the balanced loss function. We compare the performance of the optimal Bayes estimator with ones of the classical sample mean and the usual Bayes estimator under the squared error loss with respect to the posterior expected losses, risks and Bayes risks when the underlying distribution is normal as well as when they are binomial and Poisson.