• ETRI Journal
• /
• 제41권6호
• /
• pp.771-781
• /
• 2019

We propose the VS³-NET model to solve the task of question answering questions with machine-reading comprehension that searches for an appropriate answer in a given context. VS³-NET is a model that trains latent variables for each question using variational inferences based on a model of a simple recurrent unit-based sentences and self-matching networks. The types of questions vary, and the answers depend on the type of question. To perform efficient inference and learning, we introduce neural question-type models to approximate the prior and posterior distributions of the latent variables, and we use these approximated distributions to optimize a reparameterized variational lower bound. The context given in machine-reading comprehension usually comprises several sentences, leading to performance degradation caused by context length. Therefore, we model a hierarchical structure using sentence encoding, in which as the context becomes longer, the performance degrades. Experimental results show that the proposed VS³-NET model has an exact-match score of 76.8% and an F1 score of 84.5% on the SQuAD test set.

• Communications for Statistical Applications and Methods
• /
• 제30권6호
• /
• pp.589-603
• /
• 2023

Deep generative models target to infer the underlying true data distribution, and it leads to a huge success in generating fake-but-realistic data. Regarding such a perspective, the data attributes can be a crucial factor in the data generation process since non-existent counterfactual samples can be generated by altering certain factors. For example, we can generate new portrait images by flipping the gender attribute or altering the hair color attributes. This paper proposes counterfactual disentangled variational autoencoder generative adversarial networks (CDVAE-GAN), specialized for data attribute level counterfactual data generation. The structure of the proposed CDVAE-GAN consists of variational autoencoders and generative adversarial networks. Specifically, we adopt a Gaussian variational autoencoder to extract low-dimensional disentangled data features and auxiliary Bernoulli latent variables to model the data attributes separately. Also, we utilize a generative adversarial network to generate data with high fidelity. By enjoying the benefits of the variational autoencoder with the additional Bernoulli latent variables and the generative adversarial network, the proposed CDVAE-GAN can control the data attributes, and it enables producing counterfactual data. Our experimental result on the CelebA dataset qualitatively shows that the generated samples from CDVAE-GAN are realistic. Also, the quantitative results support that the proposed model can produce data that can deceive other machine learning classifiers with the altered data attributes.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 2008년도 추계학술대회 논문집
• /
• pp.263-266
• /
• 2008

This paper is concerned with the analysis of elasto-plastic stress waves in a mode I semi-infinite cracked solid subjected to Heaviside pulse load. This study adopts a time-discontinuous variational integrator based on Hamiltonian in order to reduce the numerical dispersive and dissipative errors. This also utilizes an integration scheme of the constitutive model with 2nd-order accuracy which is formulated on the strain space for a rate and temperature dependent material model. Finite element analyses of elasto-plastic stress waves are carried out in order to compare the accuracy between a conventional Galerkin method and the time- discontinuous variational integrator.

• 한국멀티미디어학회논문지
• /
• 제23권7호
• /
• pp.883-890
• /
• 2020

In this paper, we propose a variational expectation-maximization algorithm that computes posterior probabilities from Latent Dirichlet Allocation (LDA) model. The algorithm approximates the intractable posterior distribution of a document term matrix generated from a corpus made up by 50 papers. It approximates the posterior by searching the local optima using lower bound of the true posterior distribution. Moreover, it maximizes the lower bound of the log-likelihood of the true posterior by minimizing the relative entropy of the prior and the posterior distribution known as KL-Divergence. The experimental results indicate that documents clustered to image classification and segmentation are correlated at 0.79 while those clustered to object detection and image segmentation are highly correlated at 0.96. The proposed variational inference algorithm performs efficiently and faster than Gibbs sampling at a computational time of 0.029s.

• 대한수학회지
• /
• 제53권1호
• /
• pp.161-185
• /
• 2016

We consider a mathematical model which describes the quasistatic frictional contact between a piezoelectric body and an electrically conductive obstacle, the so-called foundation. A nonlinear electro-viscoelastic constitutive law is used to model the piezoelectric material. Contact is described with Signorini's conditions and a version of Coulomb's law of dry friction in which the adhesion of contact surfaces is taken into account. The evolution of the bonding field is described by a first order differential equation. We derive a variational formulation for the model, in the form of a system for the displacements, the electric potential and the adhesion. Under a smallness assumption which involves only the electrical data of the problem, we prove the existence of a unique weak solution of the model. The proof is based on arguments of time-dependent quasi-variational inequalities, differential equations and Banach's fixed point theorem.

• 대한교통학회지
• /
• 제20권3호
• /
• pp.129-147
• /
• 2002

본 연구에서 변동등식에 근거한 동적경로선택조건을 도출하여 동적통행배정모형을 제안한다. 동적경로선택조건은 운전자에 의해 이용된 경로, 링크 그리고 출발지의 출발시간을 고려하여 도출되며 이 조건을 동적통행배정모형으로 전환하는 과정에서 모형의 변동등식문제로 압축된다. 등식이론에 근거한 모델의 이론적 배경을 입증하기 위해 제안된 동적통행배정모형이 필요충분조건을 만족함이 증명되었다. 해법으로서 기존의 제안된 네트워크의 시간과 공간확장기법을 채택하지 않고 물리적 네트워크가 직접 알고리즘에 반영되도록 하기 위해 각 링크의 시간대별 통행량과 방출통행량을 링크진입통행으로 표현하여 시간종속 통행시간함수를 단일변수로 처리하여 대각화알고리즘에 반영하였다. 소규모 비대칭 네트워크 적용결과 사용자 동적최적경로선택조건이 만족됨을 입증하였는데 단위시간간격이 적을수록 개선된 효과를 보여준다. I-394네트워크 실험결과로서 기존의 변동부등식에 근거한 알고리즘에 비해 제안된 알고리즘이 최소한 93%이상의 컴퓨터연산 속도의 개선효과를 가져왔다. 등식이론에 근거한 모델개발의 장점으로서는 제안된 모델의 최적해의 계산시간이 전체시간의 증가에 전혀 영향을 받지 않는다는 것으로 이는 동적통행배정모형에 적용될 네트워크의 규모가 커질수록 제안된 알고리즘의 계산 효율성은 더욱 증가하는 것을 의미한다. 따라서 제안된 동적통행배정모형은 대규모 시간종속적 교통망에서 교통상황의 변화에 민감하게 반응할 수 있는 실시간첨단교통제어의 핵심기능으로서 역할수행이 기대된다.

• Structural Engineering and Mechanics
• /
• 제23권4호
• /
• pp.431-447
• /
• 2006

A new incremental finite element model is developed to simulate the frictional contact of elastic bodies. The incremental convex programming method is exploited, in the framework of finite element approach, to recast the variational inequality principle of contact problem in a discretized form. The non-classical friction model of Oden and Pires is adopted, however, the friction effect is represented by an equivalent non-linear stiffness rather than additional constraints. Different parametric studies are worked out to address the versatility of the proposed model.

• 한국기계가공학회지
• /
• 제7권1호
• /
• pp.38-46
• /
• 2008

Design technology and speciality production technology to manufacture high quality valve are insufficient in Korea. In order to design the experiments using Taguchi method and Variational Technology Also, from verification of the response model with optimized results was confirmed that usefulness and reliance of application Taguchi method and Variational Technology to structural's optimum design using Taguchi method and Variational Technology.

• 대한기계학회논문집A
• /
• 제28권9호
• /
• pp.1255-1261
• /
• 2004

A variational approximation procedure is introduced to study the contact stresses between a representative asperity and a feature generally happening in superfinishing processes such as FAM. After a description of the model under consideration is presented, a system of governing equation for the model is derived fullowed by the assumptions made in order to make progress in model development. Final computation is made to evaluate contact stresses on an elastic asperity tip in small scale in size and a computer simulation is performed for detailed surface profile variations on a representative feature. Numerical results are presented along with a discussion of the conclusions that can be drawn from this analysis.

• 응용통계연구
• /
• 제34권1호
• /
• pp.9-23
• /
• 2021

본 논문에서는 평균장 방법(mean-field methods)을 기반으로 사후 분포(posterior distribution)를 근사하는 방법인 변분 베이즈 방법(variational Bayes methods)에 대해 소개한다. 특히, 모수들을 실수공간으로 변환 후의 결합 사후분포를 가우시안 분포(Gaussian distribution)들의 곱(product)으로 근사하는 방법인 자동 미분 변분 추론(automatic differentiation variational inference)방법에 대해 자세히 소개하고, 환자에게 약물을 투여한 후 시간에 따라 약물의 흐름을 파악하는 연구인 약물동태학 모형(pharmacokinetic models)에 적용한다. 소개된 변분 베이즈 방법을 이용하여 자료분석을 실시하고 마코프 체인 몬테 카를로(Markov chain Monte Carlo)방법을 기초로한 자료분석의 결과와 비교한다. 알고리즘의 구현은 Stan을 이용한다.