• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.673-687
• /
• 2023

Convolutional neural network-based deep learning technology is the most commonly used in image identification, but it requires large-scale data for training. Therefore, application in specific fields in which data acquisition is limited, such as in the military, may be challenging. In particular, the identification of ground weapon systems is a very important mission, and high identification accuracy is required. Accordingly, various studies have been conducted to achieve high performance using small-scale data. Among them, the ensemble method, which achieves excellent performance through the prediction average of the pre-trained models, is the most representative method; however, it requires considerable time and effort to find the optimal combination of ensemble models. In addition, there is a performance limitation in the prediction results obtained by using an ensemble method. Furthermore, it is difficult to obtain the ensemble effect using models with imbalanced classification accuracies. In this paper, we propose a transfer learning-based feature fusion technique for heterogeneous models that extracts and fuses features of pre-trained heterogeneous models and finally, fine-tunes hyperparameters of the fully connected layer to improve the classification accuracy. The experimental results of this study indicate that it is possible to overcome the limitations of the existing ensemble methods by improving the classification accuracy through feature fusion between heterogeneous models based on transfer learning.

• Communications for Statistical Applications and Methods
• /
• v.6 no.2
• /
• pp.357-366
• /
• 1999

Let us assume that two different log-linear models are selected by various model selection methods. When these are non-hierarchical it is not easy to choose one of these models. In this paper the well-known Cox's statistic is applied to compare these non-hierarchical log-linear models. Since it is impossible to obtain the analytic solution about the problem we proposed a alternative method by extending Pesaran and pesaran's (1993) simulation approach. We find that the values of proposed test statistic and the estimates are very much stable with some empirical results.

• Journal of Information Processing Systems
• /
• v.16 no.4
• /
• pp.809-819
• /
• 2020

Gender prediction accuracy increases as convolutional neural network (CNN) architecture evolves. This paper compares voting and ensemble schemes to utilize the already trained five CNN models to further improve gender prediction accuracy. The majority voting usually requires odd-numbered models while the proposed softmax-based voting can utilize any number of models to improve accuracy. The ensemble of CNN models combined with one more fully-connected layer requires further tuning or training of the models combined. With experiments, it is observed that the voting or ensemble of CNN models leads to further improvement of gender prediction accuracy and that especially softmax-based voters always show better gender prediction accuracy than majority voters. Also, compared with softmax-based voters, ensemble models show a slightly better or similar accuracy with added training of the combined CNN models. Softmax-based voting can be a fast and efficient way to get better accuracy without further training since the selection of the top accuracy models among available CNN pre-trained models usually leads to similar accuracy to that of the corresponding ensemble models.

• Fashion & Textile Research Journal
• /
• v.20 no.2
• /
• pp.167-176
• /
• 2018

In this study, an eye-tracking program that can confirm a design cognition process was developed for the purpose of presenting strategic methods to create fashion images, and the program was used to identify what effects fashion models' external characteristics have on the cognition of design. The data for analysis were collected through an eyemovement tracking experiment and a survey, with the focus on the research problem that differences in models' external uniformity will lead to differences in the eye movement for perceiving models and design as well as the image sensibility. The results of the analysis are as follows. First, it was confirmed that the uniformity of model types and the simplicity/complexity of design led to differences in the eye movement directed at design and models and the gaze ratio. Consequently, it is deemed that models should be selected in consideration of the characteristics of design and the intention of planning when creating fashion images. Second, it was found that in terms of the cognition of design, external conditions of models affect design sensibility. A change in models led to a subtle difference in sensibility cognition even when the design condition did not change. Thus, not only the design but also model attributes are factors that should be considered important in fashion planning.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.7
• /
• pp.61-69
• /
• 2019

In this paper, a classification method of random graph models is proposed and it is based on centralities of the random graphs. Similarity between two random graphs is measured for the classification of random graph models. The similarity between two random graph models $G^{R_1}$ and $G^{R_2}$ is defined by the distance of $G^{R_1}$ and $G^{R_2}$, where $G^{R_2}$ is a set of random graph $G^{R_2}=\{G_1^{R_2},...,G_p^{R_2}\}$ that have the same number of nodes and edges as random graph $G^{R_1}$. The distance($G^{R_1},G^{R_2}$) is obtained by comparing centralities of $G^{R_1}$ and $G^{R_2}$. Through the computational experiments, we show that it is possible to compare random graph models regardless of the number of vertices or edges of the random graphs. Also, it is possible to identify and classify the properties of the random graph models by measuring and comparing similarities between random graph models.

• Journal of Climate Change Research
• /
• v.6 no.4
• /
• pp.389-397
• /
• 2015

The changes in the teleconnection associated with El Nin?o-Southern Oscillation (ENSO) over the East Asia and North Pacific under greenhouse warming are analyzed herein by comparing the Historical run (1970/1971~1999/2000) and the Representative Concentration Pathway (RCP) 4.5 run with 31 climate models, participated in the Coupled Model Intercomparison Project Phase 5 (CMIP5). It is found that CMIP5 models have diverse systematic errors in simulating the ENSO teleconnection pattern from model to model. Therefore, we select 21 models based on the models' performance in simulating teleconnection pattern in the present climate. It is shown that CMIP5 models tend to project an overall weaker teleconnection pattern associated with ENSO over East Asia in the future climate than that in the present climate. It can be also noted that the cyclonic flow over the North Pacific is weakened and shifted eastward. However, uncertainties for the ENSO teleconnection changes still exist, suggesting that much consistent agreements on this future teleconnections associated with ENSO should be taken in a further study.

• Asian Journal for Public Opinion Research
• /
• v.9 no.4
• /
• pp.352-367
• /
• 2021

Purpose: This research aims to examine gender preferences for men and women advertising models in Saudi advertisements. Saudi Arabia is known as one of the most gender-segregated society in the world, and it has gender-specific roles, characteristics, and behaviors that are undesirable for the other gender. Methodology: The questionnaire was developed with the help of earlier studies on perceptions towards advertising models and validated by a jury of experts and focus groups. The gender preferences for ten product categories (including automobiles, baby care products, cigarettes, cosmetics for women, fashion, food & beverages, motorcycles, personal care for men, personal care for women, sporting goods) were examined for men and women models. Similarly, three personal preferences characteristics for both genders (face beauty, voice quality, and Islamic dress), two characteristics for women models (body shape, femininity), and two characteristics for men models (height-weight balance, masculinity) were examined for men and women models separately. Finally, a survey was conducted to solicit responses from respondents (N=412). Findings: Results indicated significant gender preferences for gender-specific product categories and typical gender stereotypes in advertising models. Men models were preferred in men-specific products, and women models were required in women-specific products. Some product categories (including personal care for men and sporting goods) were ranked higher for men advertising models, while for women advertising models, other product categories (including personal care for women and cosmetics for women) were ranked higher. Masculinity was ranked highest as the preferred personal characteristic for men advertising models, while voice quality was highest for women advertising models. Finally, there is a significant difference between the preferred personal characteristic for men and women advertising models for three characteristics, including face beauty, Islamic dress, and masculinity and femininity. Implications: Saudi Arabia is a unique society with predominantly unique cultural dominance. Consequently, local culture greatly influences advertisements. It has stereotyped gender roles even in advertisements. This study will establish a baseline for further research on the subject area.

• Science of Emotion and Sensibility
• /
• v.10 no.1
• /
• pp.23-36
• /
• 2007

This research focused on eliciting and analyzing mental models of mobile device purchasing consumers who are distinguished by their familiarity with information technology. Mental model elicitation processes proceeded by critical decision method. And Pathfinder algorithm and Social Network Analysis were used to analyze the mental models. The results show that IT-familiar consumers have mental models of which elements are more organized and distinctive while IT-unfamiliar consumers have vague and socially affected mental models.

• Journal of Korean Society of Transportation
• /
• v.17 no.1
• /
• pp.159-171
• /
• 1999

This Paper aims to develop container terminal simulation models for medium and long term decision makings. It first undertakes in-depth survey of literature. finds its shortcomings and suggests some directions for improvement. It then proposes detailed design for the simulation models. Based on this it finally developes several simulation models and applies them to a hypothetical situation of a container terminal development. The results reveal that basic design questions such as length of quay, number of quay crane, size of storage area are well produced through the models.

• Journal of Korean Society for Geospatial Information Science
• /
• v.19 no.3
• /
• pp.83-90
• /
• 2011

The purpose of this study is to develop a series of trend analysis models for industrial water consumption and to propose a blueprint for the integration of the developed models with GIS. For the consumption data acquisition, a real-time sensing technique was adopted. Data were transformed from the field equipments to the management server in every 5 minutes. The data acquired were substituted to a polynomial formula selected. As a result, a series of models were developed for the consumption of each day. A series of validation processes were applied to the developed models and the models were finalized. Then the finalized models were transformed to the average models representing a day's average consumption or an average daily consumption of each month. Demand pattern analyses were fulfilled through the visualization of the finally derived models. It has founded out that the demand patterns show great consistency and, therefore, it is concluded that high probability of demand forecasting for a day or for a season is available. Also proposed is the integration with GIS as an IT tool by which the developed forecasting models are utilized.