• Journal of the Korean Data and Information Science Society
• /
• v.23 no.2
• /
• pp.385-392
• /
• 2012

Logistic regression is a well known binary classification method in the field of statistical learning. Mixed-effect regression models are widely used for the analysis of correlated data such as those found in longitudinal studies. We consider kernel extensions with semiparametric fixed effects and parametric random effects for the logistic regression. The estimation is performed through the penalized likelihood method based on kernel trick, and our focus is on the efficient computation and the effective hyperparameter selection. For the selection of optimal hyperparameters, cross-validation techniques are employed. Numerical results are then presented to indicate the performance of the proposed procedure.

• Communications for Statistical Applications and Methods
• /
• v.17 no.2
• /
• pp.183-191
• /
• 2010

Support vector quantile regression(SVQR) is capable of providing more complete description of the linear and nonlinear relationships among random variables. In this paper we propose an iterative reweighted least squares(IRWLS) procedure to solve the problem of SVQR with a weighted quadratic loss function. Furthermore, we introduce the generalized approximate cross validation function to select the hyperparameters which affect the performance of SVQR. Experimental results are then presented which illustrate the performance of the IRWLS procedure for SVQR.

• The Journal of Engineering Research
• /
• v.4 no.1
• /
• pp.47-54
• /
• 2002

We consider the problem of determining smoothing parameters of Gibbs priors for Bayesian methods used in the medical imaging application of emission tomographic reconstruction. We address a simple smoothing prior (membrane) whose global hyperparameter (the smoothing parameter) controls the bias/variance tradeoff of the solution. We base our maximum-likelihood(ML) estimates of hyperparameters on observed training data, and argue the motivation for this approach. Good results are obtained with a simple ML estimate of the smoothing parameter for the membrane prior.

• Journal of Korean Society for Quality Management
• /
• v.26 no.1
• /
• pp.143-160
• /
• 1998

This article is concerned with the selection of subsets of predictor variables to be included in bulding the binary response logistic regression model. It is based on a Bayesian aproach, intended to propose and develop a procedure that uses probabilistic considerations for selecting promising subsets. This procedure reformulates the logistic regression setup in a hierarchical normal mixture model by introducing a set of hyperparameters that will be used to identify subset choices. It is done by use of the fact that cdf of logistic distribution is a, pp.oximately equivalent to that of $t_{(8)}$/.634 distribution. The a, pp.opriate posterior probability of each subset of predictor variables is obtained by the Gibbs sampler, which samples indirectly from the multinomial posterior distribution on the set of possible subset choices. Thus, in this procedure, the most promising subset of predictors can be identified as that with highest posterior probability. To highlight the merit of this procedure a couple of illustrative numerical examples are given.

• ETRI Journal
• /
• v.41 no.3
• /
• pp.298-307
• /
• 2019

This paper investigates the use of the inverse-free sparse Bayesian learning (SBL) approach for peak-to-average power ratio (PAPR) reduction in orthogonal frequency-division multiplexing (OFDM)-based multiuser massive multiple-input multiple-output (MIMO) systems. The Bayesian inference method employs a truncated Gaussian mixture prior for the sought-after low-PAPR signal. To learn the prior signal, associated hyperparameters and underlying statistical parameters, we use the variational expectation-maximization (EM) iterative algorithm. The matrix inversion involved in the expectation step (E-step) is averted by invoking a relaxed evidence lower bound (relaxed-ELBO). The resulting inverse-free SBL algorithm has a much lower complexity than the standard SBL algorithm. Numerical experiments confirm the substantial improvement over existing methods in terms of PAPR reduction for different MIMO configurations.

• Journal of Korea Multimedia Society
• /
• v.24 no.4
• /
• pp.593-600
• /
• 2021

In this paper, we present the FlappyBird Competition System (FCS) implementation, a competition-based automated assessment system used in an entry-level artificial intelligence (AI) course at a university. The proposed system provides an evaluation method suitable for AI courses while taking advantage of automated assessment methods. Students are to design a neural network structure, train the weights, and tune hyperparameters using the given reinforcement learning code to improve the overall performance of game AI. Students participate using the resulting trained model during the competition, and the system automatically calculates the final score based on the ranking. The user evaluation conducted after the semester ends shows that our competition-based automated assessment system promotes active participation and inspires students to be interested and motivated to learn AI. Using FCS, the instructor significantly reduces the amount of time required for assessment.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.4
• /
• pp.126-129
• /
• 2020

We introduce a lifetime assessment technique using deep learning algorithm with complex electrical parameters such as resistivity, permittivity, impedance parameters as integrated indicators for predicting the degradation of the organic molecules. The evaluation system consists of fully automated in-situ measurement system and multiple layer perceptron learning system with five hidden layers and 1011 perceptra in each layer. Prediction accuracies are calculated and compared depending on the physical feature, learning hyperparameters. 62.5% of full time-series data are used for training and its prediction accuracy is estimated as r-square value of 0.99. Remaining 37.5% of the data are used for testing with prediction accuracy of 0.95. With k-fold cross-validation, the stability to the instantaneous changes in the measured data is also improved.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.466-469
• /
• 2021

최근 암호화폐가 발전함에 따라 다양한 연구들이 진행되고 있지만 그 중에서도 암호화폐의 가격 예측 연구들이 활발히 진행되고 있다. 특히 이러한 예측 분야에서도 인공지능 기술을 접목시켜 암호화폐 가격의 예측 정확도를 높이려는 노력들이 지속되고 있다. 인공지능 기반의 기법들 중 시간적 정보를 가진 데이터를 기반으로 하고 있는 LSTM(Long Short-Term Memory) 모델이 다각도로 활용되고 있으나 급등락하는 암호화폐 가격 데이터가 많을 경우에는 그 성능이 상대적으로 낮아질 수 밖에 없다. 따라서 본 논문에서는 가격이 급등락하고 있는 Bitcoin, Ethereum, Dash 암호화폐 데이터 환경에서 LSTM 모델의 예측 성능이 향상될 수 있는 세부 하이퍼 파라미터 값을 실험 및 분석하고, 그 결과의 의미에 대해 고찰한다. 이를 위해 LSTM 모델에서 향상된 예측률을 보일 수 있는 epoch, hidden layer 수, optimizer 에 대해 분석하였고, 최적의 예측 결과를 도출해 줄 수 있는 최소 training data 개수도 함께 살펴보았다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.657-659
• /
• 2023

최근 스마트폰과 같은 디지털 기기의 보급과 함께 개인화, 맞춤형 서비스의 수요가 늘어나면서 추천 서비스가 주목을 받고 있다. 세션 기반(Session based) 추천 시스템은 사용자의 아이템 선호에 따른 순서 정보를 고려한 학습 추천 모델로, 다양한 산업 분야에서 사용되고 있다. 세션 기반 추천 시스템 중 SASRec(Self-Attentive Sequential Recommendation) 모델은 MC/CNN/RNN 기반의 기존 여러 순차 모델들에 비하여 효율적인 성능을 보인다. 본 연구에서는 SASRec 모델의 하이퍼파라미터 중 배치 사이즈(Batch Size), 학습률 (Learning Rate), 히든 유닛(Hidden Unit)을 조정하여 실험함으로써 하이퍼파라미터에 의한 성능 변화를 분석하였다.

• ETRI Journal
• /
• v.45 no.4
• /
• pp.666-677
• /
• 2023

In the last few years, convolutional neural networks (CNNs) have demonstrated good performance while solving various computer vision problems. However, since CNNs exhibit high computational complexity, signal processing is performed on the server side. To reduce the computational complexity of CNNs for edge computing, a lightweight algorithm, such as a MobileNet, is proposed. Although MobileNet is lighter than other CNN models, it commonly achieves lower classification accuracy. Hence, to find a balance between complexity and accuracy, additional hyperparameters for adjusting the size of the model have recently been proposed. However, significantly increasing the number of parameters makes models dense and unsuitable for devices with limited computational resources. In this study, we propose a novel MobileNet architecture, in which the number of parameters is adaptively increased according to the importance of feature maps. We show that our proposed network achieves better classification accuracy with fewer parameters than the conventional MobileNet.