• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.6
• /
• pp.2527-2545
• /
• 2016

Measuring the similarity of given samples is a key problem of recognition, clustering, retrieval and related applications. A number of works, e.g. kernel method and metric learning, have been contributed to this problem. The challenge of similarity learning is to find a similarity robust to intra-class variance and simultaneously selective to inter-class characteristic. We observed that, the similarity measure can be improved if the data distribution and hidden semantic information are exploited in a more sophisticated way. In this paper, we propose a similarity learning approach for retrieval and recognition. The approach, termed as LDA-FEK, derives free energy kernel (FEK) from Latent Dirichlet Allocation (LDA). First, it trains LDA and constructs kernel using the parameters and variables of the trained model. Then, the unknown kernel parameters are learned by a discriminative learning approach. The main contributions of the proposed method are twofold: (1) the method is computationally efficient and scalable since the parameters in kernel are determined in a staged way; (2) the method exploits data distribution and semantic level hidden information by means of LDA. To evaluate the performance of LDA-FEK, we apply it for image retrieval over two data sets and for text categorization on four popular data sets. The results show the competitive performance of our method.

• Journal of the Korean Society of Safety
• /
• v.37 no.2
• /
• pp.18-27
• /
• 2022

Construction accidents are difficult to prevent because several different types of activities occur simultaneously. The current method of accident analysis only indicates the number of occurrences for one or two variables and accidents have not reduced as a result of safety measures that focus solely on individual variables. Even if accident data is analyzed to establish appropriate safety measures, it is difficult to derive significant results due to a large number of data variables, elements, and qualitative records. In this study, in order to simplify the analysis and approach this complex problem logically, data preprocessing techniques, such as latent class cluster analysis (LCCA) and predictor importance were used to discover the most influential variables. Finally, the correlation was analyzed using an alluvial flow diagram consisting of seven variables and fourteen elements based on accident data. The alluvial diagram analysis using reduced variables and elements enabled the identification of accident trends into four categories. The findings of this study demonstrate that complex and diverse construction accident data can yield relevant analysis results, assisting in the prevention of accidents.

• Korean Journal of Social Welfare
• /
• v.66 no.1
• /
• pp.307-331
• /
• 2014

The purpose of the present study was to explore latent trajectory classes in the longitudinal change of depression measured using CES-D. The study data was extracted from the Korea Welfare Panel Study Data collected from 2006 to 2010. It consisted of 8,900 adults with aged over 19. Growth Mixture Modeling(GMM) was used to explore possible latent trajectory classes in the change of depression over time. The major findings of the present study were as follows. First, there were five latent trajectory classes in the longitudinal change of depression. Second, there were 4 latent trajectory classes of depression for people in a non-poverty group, while there were 3 latent trajectory classes of depression for people in a poverty group. These findings lead to three conclusions. First, 12.1% of the sample shows that their depression level increases over time. Second, the previous research findings of decreased depression over time might be caused by the combination of two latent trajectory classes(a low level depression sustain group and a depression decrease group). Lastly, the latent trajectory classes in the longitudinal change of depression, which are found in the present study, might be caused by interactions among depression, age, and poverty status.

• Journal of rehabilitation welfare engineering & assistive technology
• /
• v.6 no.1
• /
• pp.83-87
• /
• 2012

This study is aimed to identify latent classes which are based the change patterns in assistive technology device use among worker with disabilities and to test the effects of independent variables(gender, education, disability type, disability density, activity and participation of ICF: ICF, subjective socioeconomic status: SES, job satisfaction, life satisfaction) on determining latents classes. This study applied Nagin's(1999) semi-parametric group based approach to the panel survey of employment for the disabled. Because dependant variable has dichotomous scale, logit model was used. The results identified three latent classes, which could be defined based on the patterns as follows; assistive device continued use group, assistive device mid-level use group, assistive device sharp decline use group. The effects of the independent variables on the latent classes was tested by multinomial logit analysis. The results showed that education, disability type, ICF, SES, and life satisfaction were significant determinants of the latent classes. Finally, the implications based on analysis results were suggested.

• Korean Journal of Optics and Photonics
• /
• v.31 no.3
• /
• pp.125-133
• /
• 2020

An innovative method for separating overlapping latent fingerprints, using laser-induced plasma spectroscopy (LIPS) combined with multivariate analysis, is reported in the current study. LIPS provides the capabilities of real-time analysis and high-speed scanning, as well as data regarding the chemical components of overlapping fingerprints. These spectra provide valuable chemical information for the forensic classification and reconstruction of overlapping latent fingerprints, by applying appropriate multivariate analysis. This study utilizes principal-component analysis (PCA) and partial-least-squares (PLS) techniques for the basis classification of four types of fingerprints from the LIPS spectra. The proposed method is successfully demonstrated through a classification example of four distinct latent fingerprints, using discrimination such as soft independent modeling of class analogy (SIMCA) and partial-least-squares discriminant analysis (PLS-DA). This demonstration develops an accuracy of more than 85% and is proven to be sufficiently robust. In addition, by laser-scanning analysis at a spatial interval of 125 ㎛, the overlapping fingerprints were separated as two-dimensional forms.

• IMMUNE NETWORK
• /
• v.23 no.2
• /
• pp.17.1-17.16
• /
• 2023

Latent membrane protein 2A (LMP2A), a latent Ag commonly expressed in Epstein-Barr virus (EBV)-infected host cells, is a target for adoptive T cell therapy in EBV-associated malignancies. To define whether individual human leukocyte antigen (HLA) allotypes are used preferentially in EBV-specific T lymphocyte responses, LMP2A-specific CD8⁺ and CD4⁺ T cell responses in 50 healthy donors were analyzed by ELISPOT assay using artificial Ag-presenting cells expressing a single allotype. CD8⁺ T cell responses were significantly higher than CD4⁺ T cell responses. CD8⁺ T cell responses were ranked from highest to lowest in the order HLA-A, HLA-B, and HLA-C loci, and CD4⁺ T cell responses were ranked in the order HLA-DR, HLA-DP, and HLA-DQ loci. Among the 32 HLA class I and 56 HLA class II allotypes, 6 HLA-A, 7 HLA-B, 5 HLA-C, 10 HLA-DR, 2 HLA-DQ, and 2 HLA-DP allotypes showed T cell responses higher than 50 spot-forming cells (SFCs)/5×105 CD8⁺ or CD4⁺ T cells. Twenty-nine donors (58%) showed a high T cell response to at least one allotype of HLA class I or class II, and 4 donors (8%) had a high response to both HLA class I and class II allotypes. Interestingly, we observed an inverse correlation between the proportion of LMP2A-specific T cell responses and the frequency of HLA class I and II allotypes. These data demonstrate the allele dominance of LMP2A-specific T cell responses among HLA allotypes and their intra-individual dominance in response to only a few allotypes in an individual, which may provide useful information for genetic, pathogenic, and immunotherapeutic approaches to EBV-associated diseases.

• Korean Journal of Culture and Social Issue
• /
• v.28 no.3
• /
• pp.483-512
• /
• 2022

This study examined the patterns of coping strategies among Koreans during the early stage of the COVID-19 pandemic, explored the influence of demographic information (gender, age, economic level, household type), along with the unusual experiences due to COVID-19 (fear, stress of COVID, constraints of routine, income risk) on the classification of subclasses, and analyzed the latent profile differences in psychological wellbeing (life satisfaction, depression, and anxiety). An online survey was conducted among Korean Adults(n=600) between April 13, 2020 and 21, when WHO declared COVID-19 a global pandemic and Daegu as well as Gyeongsangbuk-do was nominated as a special disaster zone. First, Latent Profile Analysis (LPA) was used to identify subclasses of coping strategies and results suggested that the 4-class model had the best fit. Second, Class memberships were predicted by gender, age, economic level, as well as fear, stress, constraints of routine, and income risk, among the unusual experiences due to COVID-19. Finally, there are differences in psychological wellbeing among latent profiles. 'High level of adaptive coping group 3' showed the highest level of life satisfaction, 'Adaptive-maladaptive coping group 4' showed the highest level of depression, anxiety. Implications and suggestions are discussed based on the study results.

• Korean Journal of Social Welfare Studies
• /
• v.40 no.3
• /
• pp.377-406
• /
• 2009

This study used laten class growth analysis to identify discrete developmental patterns of delinquent behaviors in adolescence. This present article also examined associations among these trajectories to determine how the development of delinquent behaviors relates to protective and risk factors, which include parental monitoring, attachment with parent, association with deviant peers, self-control, and negative stigma from others. Four-wave panel data from a Korea Youth Panel Study were used for the latent class growth model analysis. The sample consisted of 3,446 adolescents who were assessed at 4 measurement waves with approximately 1-year interval. Four trajectories of delinquent behaviors emerged: delinquency persistence, delinquency increaser, delinquency decreaser, normative group(almost no delinquent behaviors). Association with deviant peers had the most proximal strong influence on the probability of being in the delinquency increaser and delinquency persistence group compared, noed to the normative group. Parental monitoring, self-efficacy and negative stigma also differentiated the four delinquent behavior trajectories from one another after controllig for socio-demographic variables. The study suggested that there is a significant heterogeneity in the timing and change rate of delinquency progression. Adolescent delinquency prevention and intervention programs will need to consider this heterogeneity and enhance attention to protective and risk factors depending on the subpopulation.

• 한국노년학
• /
• v.40 no.1
• /
• pp.33-49
• /
• 2020

This study was conducted to examine the latent classes in social exclusion and to analyse empirically the effects on the economic instability of old age by this type. And it also sought to look at whether the influence of old age anxiety varies with the subjective class consciousness of the elderly. Using the 14th data from the Korea General Social Survey (KGSS) in 2016, 1,041 adult males and females aged 18 years old were analyzed at the time of the survey. T-test, potential layer analysis (LCA), and multinomantic analysis of potential groups were conducted using the STATA14 and MPLUS 7 statistical programs. Finally, multi-regression analysis was performed to identify the moderate effect and effects among variables. According to the research, the types of social exclusion were three groups, followed by social exclusion group (49.3%), Multi-dimensional exclusion group (30.9%), and active social participation group (19.7%). The social exclusion group has the lowest possibility of economic, employment, and health exclusion, but the exclusion of formal and informal social activities seem to prominent, and the multi-dimensional exclusion group is more than 50% likely to experience exclusion in all areas. Active social participation are characterized by very active participation in informal social activities. By conducting multinominal logistic regression, it was observed that the social exclusion group included more young people than other groups, and that the multi-dimensional exclusion group included many elderly women without spouses. Finally, multiple regression analysis showed that social exclusion type interacts with subjective class consciousness and affects economic anxiety of old age.

• IEIE Transactions on Smart Processing and Computing
• /
• v.4 no.4
• /
• pp.202-208
• /
• 2015

In this study, we propose a new inference algorithm for a multiclass Gaussian process classification model using a variational EM framework and the Laplace approximation (LA) technique. This is performed in two steps, called expectation and maximization. First, in the expectation step (E-step), using Bayes' theorem and the LA technique, we derive the approximate posterior distribution of the latent function, indicating the possibility that each observation belongs to a certain class in the Gaussian process classification model. In the maximization step, we compute the maximum likelihood estimators for hyper-parameters of a covariance matrix necessary to define the prior distribution of the latent function by using the posterior distribution derived in the E-step. These steps iteratively repeat until a convergence condition is satisfied. Moreover, we conducted the experiments by using synthetic data and Iris data in order to verify the performance of the proposed algorithm. Experimental results reveal that the proposed algorithm shows good performance on these datasets.