• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.609-612
• /
• 2018

기계학습 분야에서 분류 문제를 해결하기 위해 다양한 알고리즘들이 연구되고 있다. 하지만 기존에 연구된 분류 알고리즘 대부분은 각 클래스에 속한 데이터 수가 거의 같다는 가정하에 학습을 진행하기 때문에 각 클래스의 데이터 수가 불균형한 경우 분류 정확도가 다소 떨어지는 현상을 보인다. 이러한 문제를 해결하기 위해 본 논문에서는 Conditional Generative Adversarial Networks(CGAN)을 활용하여 데이터 수의 균형을 맞추는 오버샘플링 기법을 제안한다. CGAN은 데이터 수가 적은 클래스에 속한 데이터 특징을 학습하고 실제 데이터와 유사한 데이터를 생성한다. 이를 통해 클래스별 데이터의 수를 맞춰 분류 알고리즘의 분류 정확도를 높인다. 실제 수집된 데이터를 이용하여 CGAN을 활용한 오버샘플링 기법이 효과가 있음을 보이고 기존 오버샘플링 기법들과 비교하여 기존 기법들보다 우수함을 입증하였다.

• Language and Information
• /
• v.8 no.1
• /
• pp.137-161
• /
• 2004

In this paper, I will look at two conditional connectives in Korean and point out differences between -(u)myen and -ta/la-myen in their distributions and semantics. One of the differences is that -ta/la-myen always allows epistemic interpretation, whereas -(u)myen allows epistemic interpretation only when the event time of the antecedent clause is in the past or present. A second difference is that only -(u)myen is used in purely temporal and habitual conditionals. A third difference is that the modality marker -keyss, which can have volitional or predictive interpretation with -(u)myen, cannot have predictive interpretation with -ta/la-myen. I propose that -ta/la-myen has the operator of settledness, which is defined with respect to the speech time, and explain the differences listed based on the semantics of settledness.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.4
• /
• pp.1191-1200
• /
• 2006

A chi-squared test of multivariate normality is suggested which is mainly focused on detecting deviations from elliptical symmetry. This test uses Mahalanobis distances of observations to have some power for deviations from multivariate normality. We derive the limiting distribution of the test statistic by a conditional limit theorem. A simulation study is conducted to study the accuracy of the limiting distribution in finite samples. Finally, we compare the power of our method with those of other popular tests of multivariate normality under two non-normal distributions.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.2
• /
• pp.344-352
• /
• 1997

This paper proposes an efficient lossless coding scheme for tree structured vector quantization (TSVQ) system which efficiently exploits inter-block correlation. The TSVQ index of the current block is adaptively arithmeticencoded depending on the indices of the previous blocks. This paper also presents a reductio method, which effectively resolve the memory problem which usually arises in many conditional entropy coding schemes. Simulation results show that the proposed scheme provides remarkable bitrate reduction by effectively exploiting not only linear but also non-linear inter-block correlation.

• Journal of the Korean Data and Information Science Society
• /
• v.6 no.2
• /
• pp.9-15
• /
• 1995

In discriminant analysis the procedures commonly used to estimate the dimensionality involve testing a sequence of dimensionality hypotheses. There is a problem with the size of the test since dimensionality hypotheses are tested sequentially and thus they are actually conditional tests. The focus of this paper is "How is the size of the test affected by viewing this sequence of tests as conditional tests?".

• Journal of Communications and Networks
• /
• v.13 no.6
• /
• pp.621-624
• /
• 2011

With the rapid growth of internet traffic, accurate and reliable prediction of internet traffic has been a key issue in network management and planning. This paper proposes an autoregressive-generalized autoregressive conditional heteroscedasticity (AR-GARCH) error model for forecasting internet traffic and evaluates its performance by comparing it with seasonal autoregressive integrated moving average (ARIMA) models in terms of root mean square error (RMSE) criterion. The results indicated that the seasonal AR-GARCH models outperformed the seasonal ARIMA models in terms of forecasting accuracy with respect to the RMSE criterion.

• Journal of the Korean Data and Information Science Society
• /
• v.10 no.1
• /
• pp.207-214
• /
• 1999

Optimal estimating functions for a class of autoregressive models with GARCH(1,1) error are discussed. The asymptotic properties of the estimator as the solution of the optimal estimating equation are investigated for the models. We have also some simulation results which suggest that the proposed optimal estimators have smaller sample variances than those of the Conditional least-squares estimators under the heavy-tailed error distributions.

• Journal of the Korean Data and Information Science Society
• /
• v.25 no.5
• /
• pp.1107-1116
• /
• 2014

In this paper, we review recent developments in network analysis using the graph theory, and introduce ongoing research area with relevant theoretical results. In specific, we introduce basic notations in graph, and conditional and marginal approach in constructing the adjacency matrix. Also, we introduce the Marcenko-Pastur law, the Tracy-Widom law, the white Wishart distribution, and the spiked distribution. Finally, we mention the relationship between degrees and eigenvalues for the detection of hubs in a network.

• Language and Information
• /
• v.6 no.1
• /
• pp.21-40
• /
• 2002

This paper explores Discourse Rep-resentation Structures which can successfully describe the mental representations that discourse participants form when they hear so-called double access sentences. The syntactic, semantic and pragmatic characteristics of double access sentences are discussed. The analysis proposed in this paper, employing a modified version of the 'conditional beliefs' of Chung(1997), successfully explains the semantic and pragmatic characteristics of present or future tense in double access sentences as well as when and why the speaker should take or can be exempted from the responsibility for using present or future tense in double access sentences.

• Journal of Families and Better Life
• /
• v.26 no.5
• /
• pp.333-354
• /
• 2008

This study aimed to investigate the role of domain-specific causal mechanism information and domain-general conditional probability in young children's causal reasoning on psychology and biology. Participants were 121 3-year-olds and 121 4-year-olds recruited from seven childcare centers in Seoul, Kyonggi Province, and Busan. After participants watched moving pictures on psychological and biological phenomena, they were asked to choose appropriate cause and justify their choices. Results of this study were as follows: First, young children made different inferences according to domain-specific causal mechanisms. Second, the developmental level of causal mechanisms has a gap between psychology and biology, and biological knowledge was proved to be separate from psychological knowledge during the preschool period. Third, young children's causal reasoning was different depending on the interaction effect of domain-specific mechanisms and domain-general conditional probability: children could make more inferences based on domain-specific causal mechanisms if conditional probability between domain-appropriate cause and effect was evident. To conclude, it can be inferred that the role of domain-specific causal mechanisms and domain-general conditional probability is not competitive but complementary in young children's causal reasoning.