• Smart Media Journal
• /
• v.9 no.3
• /
• pp.80-89
• /
• 2020

Out - of - distribution (OOD) samples are frequently encountered when deploying a classification model in plenty of real-world machine learning-based applications. Those samples are normally sampling far away from the training distribution, but many classifiers still assign them high reliability to belong to one of the training categories. In this study, we address the problem of removing OOD examples by estimating marginal density estimation using variational autoencoder (VAE). We also investigate other proper methods, such as temperature scaling, Gaussian discrimination analysis, and label smoothing. We use Chonnam National University (CNU) weeds dataset as the in - distribution dataset and CIFAR-10, CalTeach as the OOD datasets. Quantitative results show that the proposed framework can reject the OOD test samples with a suitable threshold.

• Toxicological Research
• /
• v.22 no.4
• /
• pp.423-429
• /
• 2006

Recent reports on the photocarcinogenicity and photogerotoxicity of many compounds led to an increasing awareness for the need of a standard approach to test for photogenotoxicity. The comet assay has been recently validated as a sensitive and specific test system for the quantification of DNA damage. Thus, the objectives of this study are to investigate the utility of photo-comet assay for detecting photo-mutagens, and to evaluate its ability to predict rodent photo-carcinogenicity. Photo-comet assays were performed using L5178Y $Tk^{+/-}$ mouse lymphoma cells on five test substances (8-methoxypsoralen, chlorpromazine, lomefloxacin, anthracene and retinoic acid) that demonstrated positive results in photocarcinogenicity tests. For the best discrimination between the test substance-mediated DNA damage and the undesirable DNA damage caused by direct UV absorption, a UV dose-response of the cells in the absence of the test substances was firstly fnalized. Out of 5 test substances, positive comet results were obtained for chlorpromazine, lomefloxacin, anthracene and retinoic acid while 8-methoxypsoralen found negative. An investigation into the predictive value of this photo-comet assay for determining the photocarcinogenicity showed that photo-comet assay has relatively high sensitivity. Therefore, the photo-comet assay with mammalian cells seems to be a good and sensitive predictor of the photocarcinogenic potential of new substances.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.3
• /
• pp.522-538
• /
• 2013

Content based image retrieval has become an increasingly important research topic for its wide application. It is highly challenging when facing to large-scale database with large variance. The retrieval systems rely on a key component, the predefined or learned similarity measures over images. We note that, the similarity measures can be potential improved if the data distribution information is exploited using a more sophisticated way. In this paper, we propose a similarity measure learning approach for image retrieval. The similarity measure, so called Fisher kernel, is derived from the probabilistic distribution of images and is the function over observed data, hidden variable and model parameters, where the hidden variables encode high level information which are powerful in discrimination and are failed to be exploited in previous methods. We further propose a discriminative learning method for the similarity measure, i.e., encouraging the learned similarity to take a large value for a pair of images with the same label and to take a small value for a pair of images with distinct labels. The learned similarity measure, fully exploiting the data distribution, is well adapted to dataset and would improve the retrieval system. We evaluate the proposed method on Corel-1000, Corel5k, Caltech101 and MIRFlickr 25,000 databases. The results show the competitive performance of the proposed method.

• Nuclear Engineering and Technology
• /
• v.46 no.6
• /
• pp.725-736
• /
• 2014

It is very important to detect and identify small anomalies and component failures for the safe operation of complex and large-scale artifacts such as nuclear power plants. Each diagnostic technique has its own advantages and limitations. These facts inspire us not only to enhance the capability of diagnostic techniques but also to integrate the results of diagnostic subsystems in order to obtain more accurate diagnostic results. The article describes the outline of four diagnostic techniques developed for the condition monitoring of the fast breeder reactor "Monju". The techniques are (1) estimation technique of important state variables based on a physical model of the component, (2) a state identification technique by non-linear discrimination function applying SVM (Support Vector Machine), (3) a diagnostic technique applying WT (Wavelet Transformation) to detect changes in the characteristics of measurement signals, and (4) a state identification technique effectively using past cases. In addition, a hybrid diagnostic system in which a final diagnostic result is given by integrating the results from subsystems is introduced, where two sets of values called confidence values and trust values are used. A technique to determine the trust value is investigated under the condition that the confidence value is determined by each subsystem.

• Journal of Fisheries and Marine Sciences Education
• /
• v.28 no.4
• /
• pp.1076-1088
• /
• 2016

This study was to validate the WSBS_PECT (Workplace Spirituality Belief Scale for Prospective Early Childhood Teacher) using discriminant analysis on prospective early childhood teachers' happiness and career maturity. The data from 523 prospective early childhood teachers were analyzed statistically through t-test and binary logistic regression model. The results indicated that 1) the higher group in workplace spirituality belief significantly gets more scores of happiness and career maturity than the lower group, 2) 1 factors of the WSBS_PECT has discriminant power on prospective early childhood teachers' happiness, and 3) 2 factors ('meaning for life' and 'belief on calling for ECE teacher job') of the WSBS_PECT are effective to discriminate prospective early childhood teachers' career maturity. Further statistical works are supplementary needed to validate the WSBS_PECT and to increase its' feasibility.

• Journal of Korean Academy of Nursing
• /
• v.46 no.1
• /
• pp.109-117
• /
• 2016

Purpose: The aim of this study was to develop a Korean version of Night Eating Questionnaire (KNEQ) and test its psychometric properties and evaluate items according to item response theory. Methods: The 14-item NEQ as a measure of severity of the night eating syndrome was translated into Korean, and then this KNEQ was evaluated. A total of 1171 participants aged 20 to 50 completed the KNEQ on the Internet. To test reliability and validity, Cronbach's alpha, correlation, simple regression, and factor analysis were used. Each item was analyzed according to Rasch-Andrich rating scale model and item difficulty, discrimination, infit/outfit, and point measure correlation were evaluated. Results: Construct validity was evident. Cronbach's alpha was .78. The items of evening hyperphagia and nocturnal ingestion showed high ability in discriminating people with night eating syndrome, while items of morning anorexia and mood/sleep provided relatively little information. The results of item analysis showed that item2 and item7 needed to be revised to improve the reliability of KNEQ. Conclusion: KNEQ is an appropriate instrument to measure severity of night eating syndrome with good validity and reliability. However, further studies are needed to find cut-off scores to screen persons with night eating syndrome.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.52 no.4
• /
• pp.179-188
• /
• 2003

Percentage differential characteristic scheme has been recognized as the principal basis for large transformer protection. Nowadays, relaying signals can contain second harmonic component to a large extent even in a normal state, and second harmonic ratio indicates a tendency of relative reduction because of the advancement of transformer's core material. And then, conventional second harmonic restraint differential relaying exposes some doubt in reliability. It is, therefore, necessary to develop a new algorithm for the effective and accurate discrimination. This paper deals with advanced fuzzy logic based relaying by using flux differential, and a new fault detection criterion logic scheme by using wavelet transform. To comparative analysis of proposed techniques, the paper constructs power system model including power transformer, utilizing the EMTP, and collects data through simulation of various internal faults and magnetizing inrush. The proposed fuzzy relaying and a new fault detection scheme were tested. The former, fuzzy relaying, was proven to be faster and more reliable than the latter.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.61 no.12
• /
• pp.1801-1806
• /
• 2012

This paper presents a simple and practical method which enables to prevent malfunction of protection relay due to differential current caused by one end CT saturation in case of external fault. This method uses difference of magnitude(instantaneous value) between the both end current just before the occurrence of differential current without a separate method to CT staturation detection. One end CT saturation is simulated by current transformer model using type-96 component and the presented method is verified by using EMTP MODELS with respect to internal and external fault with one end CT staturation. The presented method distinguished rightly bewteen external and internal fault with one end CT saturation. This information can be used to prevent malfunction of current differential protection relay in case of external fault. And this method is not affected by sampling rate and has no calculation burden, so it will be applicable to differential current protection relay with ease.

• Journal of the Ergonomics Society of Korea
• /
• v.18 no.1
• /
• pp.13-24
• /
• 1999

The objective of the study is to discriminate EEG(electroencephalogram) due to emotional changes. Emotion was evoked by the series of auditory stimuli which were selected from the natural sounds in the sound effect collection of compact disc. Seventeen university students participated and experienced positive or negative emotions by six auditory stimuli with intermission between stimuli. Temporal EEG ($T_3$, $T_4$, $T_5$, and $T_6$) was recorded at the same time and a subjective test was performed on the eleven point scales after the experiment. The maximum and minimum scores of the EEG among six stimuli EEG were analyzed for discrimination of emotion. The EEG signals were transformed into feature objects based on scalar intervention model coefficients. Auditory stimulus was considered as intervention variable. They were classified by Discriminant Analysis for each channel. The features showed results with the best classification accuracy of 91.2 % in $T_4$ for auditory stimuli. This study could be extended to establish an algorithm which quantifies and classifies emotions evoked by auditory stimulus using time-series models.

• Phonetics and Speech Sciences
• /
• v.2 no.2
• /
• pp.63-68
• /
• 2010

A new version of sequential probability ratio test (SPRT) which has been investigated in utterance-length control is proposed to obtain uniform response results in speaker verification (SV). Although SPRTs can obtain fast responses in SV tests, differences in the performance may occur depending on the compositions of consonants and vowels in the sentences used. In this paper, a fast sequential probability ratio test (FSPRT) method that shows consistent performances at all times regardless of the compositions of vocalized sentences for SV will be proposed. In generating frames, the FSPRT will first conduct SV test processes with only generated frames without any overlapping and if the results do not satisfy discrimination criteria, the FSPRT will sequentially use frames applied with overlapping. With the progress of processes as such, the test will not be affected by the compositions of sentences for SV and thus fast response outcomes and even consistent performances can be obtained. Experimental results show that the FSPRT has better performance to the SPRT method while requiring less complexity with equal error rates (EER).