• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.352-357
• /
• 2023

자연어 추론은 두 문장(전제, 가설)간의 관계를 이해하고 추론하여 함의, 모순, 중립 세 가지 범주로 분류하며, 전제-가설-라벨(PHL) 데이터셋을 활용하여 자연어 추론 모델을 학습한다. 그러나, 새로운 도메인에 자연어 추론을 적용할 경우 학습 데이터가 존재하지 않거나 이를 구축하는 데 많은 시간과 자원이 필요하다는 문제가 있다. 본 논문에서는 자연어 추론을 위한 학습 데이터인 전제-가설-라벨 삼중항을 자동 생성하기 위해 [1]에서 제안한 문장 변환 규칙 대신에 거대 언어 모델과 Chain-of-Thought(CoT), Program-aided Language Models(PaL) 등의 프롬프팅(Prompting) 방법을 이용하여 전제-가설-라벨 삼중항을 자동으로 생성하는 방법을 제안한다. 실험 결과, CoT와 PaL 프롬프팅 방법으로 자동 생성된 데이터의 품질이 기존 규칙이나 기본 프롬프팅 방법보다 더 우수하였다.

• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.200-207
• /
• 2020

Recently, deep neural networks such as CNN are showing excellent performance in various fields such as image classification, object recognition, visual quality enhancement, etc. However, as the model size and computational complexity of deep learning models for most applications increases, it is hard to apply neural networks to IoT and mobile environments. Therefore, neural network compression algorithms for reducing the model size while keeping the performance have been being studied. In this paper, we apply few compression methods to CNN models and evaluate their performances in the embedded environment. For evaluate the performance, the classification performance and inference time of the original CNN models and the compressed CNN models on the image inputted by the camera are evaluated in the embedded board equipped with QCS605, which is a customized AI chip. In this paper, a few CNN models of MobileNetV2, ResNet50, and VGG-16 are compressed by applying the methods of pruning and matrix decomposition. The experimental results show that the compressed models give not only the model size reduction of 1.3~11.2 times at a classification performance loss of less than 2% compared to the original model, but also the inference time reduction of 1.2~2.21 times, and the memory reduction of 1.2~3.8 times in the embedded board.

• Structural Engineering and Mechanics
• /
• v.47 no.3
• /
• pp.401-419
• /
• 2013

In this study, the efficiency of adaptive neuro-fuzzy inference system (ANFIS) and genetic expression programming (GEP) in predicting the effects of infill walls on base reactions and roof drift of reinforced concrete frames were investigated. Current standards generally consider weight and fundamental period of structures in predicting base reactions and roof drift of structures by neglecting numbers of floors, bays, shear walls and infilled bays. Number of stories, number of bays in x and y directions, ratio of shear wall areas to the floor area, ratio of bays with infilled walls to total number bays and existence of open story were selected as parameters in GEP and ANFIS modeling. GEP and ANFIS have been widely used as alternative approaches to model complex systems. The effects of these parameters on base reactions and roof drift of RC frames were studied using 3D finite element method on 216 building models. Results obtained from 3D FEM models were used to in training and testing ANFIS and GEP models. In ANFIS and GEP models, number of floors, number of bays, ratio of shear walls and ratio of infilled bays were selected as input parameters, and base reactions and roof drifts were selected as output parameters. Results showed that the ANFIS and GEP models are capable of accurately predicting the base reactions and roof drifts of RC frames used in the training and testing phase of the study. The GEP model results better prediction compared to ANFIS model.

• Communications for Statistical Applications and Methods
• /
• v.24 no.6
• /
• pp.605-625
• /
• 2017

This paper presents proportional odds cure models to allow spatial correlations by including spatial frailty in the interval censored data setting. Parametric cure rate models with independent and dependent spatial frailties are proposed and compared. Our approach enables different underlying activation mechanisms that lead to the event of interest; in addition, the number of competing causes which may be responsible for the occurrence of the event of interest follows a Geometric distribution. Markov chain Monte Carlo method is used in a Bayesian framework for inferential purposes. For model comparison some Bayesian criteria were used. An influence diagnostic analysis was conducted to detect possible influential or extreme observations that may cause distortions on the results of the analysis. Finally, the proposed models are applied for the analysis of a real data set on smoking cessation. The results of the application show that the parametric cure model with frailties under the first activation scheme has better findings.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.12
• /
• pp.519-524
• /
• 2023

In hashing-based image retrieval, the hash code of a manipulated image is different from the original image, making it difficult to search for the same image. This paper proposes and evaluates a self-supervised deephashing model that generates perceptual hash codes from feature information such as texture, shape, and color of images. The comparison models are autoencoder-based variational inference models, but the encoder is designed with a fully connected layer, convolutional neural network, and transformer modules. The proposed model is a variational inference model that includes a SimAM module of extracting geometric patterns and positional relationships within images. The SimAM module can learn latent vectors highlighting objects or local regions through an energy function using the activation values of neurons and surrounding neurons. The proposed method is a representation learning model that can generate low-dimensional latent vectors from high-dimensional input images, and the latent vectors are binarized into distinguishable hash code. From the experimental results on public datasets such as CIFAR-10, ImageNet, and NUS-WIDE, the proposed model is superior to the comparative model and analyzed to have equivalent performance to the supervised learning-based deephashing model. The proposed model can be used in application systems that require low-dimensional representation of images, such as image search or copyright image determination.

• Communications for Statistical Applications and Methods
• /
• v.22 no.6
• /
• pp.557-573
• /
• 2015

In this paper we develop a Bayesian inference for a multiplicative double seasonal autoregressive (DSAR) model by implementing a fast, easy and accurate Gibbs sampling algorithm. We apply the Gibbs sampling to approximate empirically the marginal posterior distributions after showing that the conditional posterior distribution of the model parameters and the variance are multivariate normal and inverse gamma, respectively. The proposed Bayesian methodology is illustrated using simulated examples and real-world time series data.

• Journal of the Korean Data and Information Science Society
• /
• v.5 no.2
• /
• pp.11-18
• /
• 1994

Sample sizes for experiments concerned with inference on R=Pr(X

• Communications for Statistical Applications and Methods
• /
• v.9 no.3
• /
• pp.825-834
• /
• 2002

Poisson processes are widely used in reliability and survival analysis. In particular, multiple event time data in survival analysis are routinely analyzed by use of Poisson processes. In this paper, we consider large sample properties of nonparametric Bayesian models for Poisson processes. We prove that the posterior distribution of the cumulative intensity function of Poisson processes is consistent under regularity conditions on priors which are Levy processes.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.387-390
• /
• 1998

In this paper, we suggested a neuro-fuzzy learning algorithm for tuning fuzzy rules, in which a fuzzy system model is of additive-type. Using the method, it is possible to reduce the computation size, since performing the fuzzy inference and tuning the fuzzy rules of each fuzzy subsystem model are independent. Moreover, the efficiency of suggested method is shown by means of a numerical example.

• Journal of Korean Society for Quality Management
• /
• v.32 no.4
• /
• pp.242-251
• /
• 2004

Several statistical methods are introduced 1=o analyze the accelerated failure time data. Most frequently used method is the log-linear approach with parametric assumption. Since the accelerated failure time experiments are exposed to many environmental restrictions, parametric log-linear relationship might not be working properly to analyze the resulting data. The models proposed by Buckley and James(1979) and Stute(1993) could be useful in the situation where parametric log-linear method could not be applicable. Those methods are introduced in accelerated experimental situation under the thermal acceleration and discussed through an illustrated example.