• Journal of Intelligence and Information Systems
• /
• v.28 no.3
• /
• pp.45-55
• /
• 2022

Recently, large-scale language models (LPLM) have been shown state-of-the-art performances in various tasks of natural language processing including intent classification. However, fine-tuning LPLM requires much computational cost for training and inference which is not appropriate for dialog system. In this paper, we propose compressed intent classification model for multi-agent in low-resource like CPU. Our method consists of two stages. First, we trained sentence encoder from LPLM then compressed it through knowledge distillation. Second, we trained agent-specific adapter for intent classification. The results of three intent classification datasets show that our method achieved 98% of the accuracy of LPLM with only 21% size of it.

• Physical Therapy Korea
• /
• v.12 no.4
• /
• pp.20-25
• /
• 2005

The Gross Motor Function Measure (GMFM) is an internationally widely used outcome measure. The aim of this study was to evaluate the structural properties of the Korean version of GMFM using the Rasch Model, with regard to scoring within rehabilitation centers in Korea. GMFM data for 206 children with cerebral palsy were collected from 11 outpatient rehabilitation facilities by 29 pediatric therapists. The Winsteps software was used to refine the rating scale. This study suggests that the scoring categories of the Korean version of the GMFM should be collapsed from 0 (subject does not initiate task), 1 (subject initiates task), 2 (subject partially completes task), 3 (subject completes task) to 0 (subject does not initiate task), 1 (subject initiates or partially completes task), 2 (subject completes task) for better accuracy in estimating the gross motor function of children with cerebral palsy.

• Journal of the Korea Society for Simulation
• /
• v.21 no.2
• /
• pp.59-67
• /
• 2012

A grid computing consists of the physical resources for processing one of the large-scale jobs. However, due to the recent trends of rapid growing data, the grid computing needs a parallel processing method to process the job. In general, each physical resource divides a requested large-scale task. And a processing time of the task varies with an efficiency and a distance of each resource. Even if some resource completes a job, the resource is standing by until every divided job is finished. When every resource finishes a processing, each resource starts a next job. Therefore, this paper proposes a dynamic resource reallocation scheduling model (DDRSM). DDRSM finds a waiting resource and reallocates an unfinished job with an efficiency and a distance of the resource. DDRSM is an efficient method for processing multiple large-scale jobs.

• Korean Journal of Child Studies
• /
• v.32 no.1
• /
• pp.51-69
• /
• 2011

This study examined how young children understand representational relations between referents and their representational objects. Ninety-four children aged 2- to 4.5-years of age were individually tested; firstly in the scale-model tasks, and then in the scale-map tasks. Data were analyzed both by means of Chi-Square test and by a more descriptive, micro analysis. According to the results, there were significant age differences in the understanding of representational relations, regardless of the type of representational objects. In the descriptive, micro analysis, it was found that before 3 years of age, young children have a great deal of difficulties in understanding representational relations. More importantly, young children under three seemed unable to understand representational relations, especially when the similarities as well as the differences between the representational object and the referent were very high. These results suggest that teachers of very young children need to select representational materials carefully, taking into consideration children's understanding of representational relations.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.25-28
• /
• 2020

Automatic segmentation of brain tissues such as WM, GM, and CSF from brain MRI scans is helpful for the diagnosis of many neurological disorders. Accurate segmentation of these brain structures is a very challenging task due to low tissue contrast, bias filed, and partial volume effects. With the aim to improve brain MRI segmentation accuracy, we propose an end-to-end convolutional based U-SegNet architecture designed with multi-scale kernels, which includes cascaded dilated convolutions for the task of brain MRI segmentation. The multi-scale convolution kernels are designed to extract abundant semantic features and capture context information at different scales. Further, the cascaded dilated convolution scheme helps to alleviate the vanishing gradient problem in the proposed model. Experimental outcomes indicate that the proposed architecture is superior to the traditional deep-learning methods such as Segnet, U-net, and U-Segnet and achieves high performance with an average DSC of 93% and 86% of JI value for brain MRI segmentation.

• Journal of Korean Academy of Nursing
• /
• v.27 no.4
• /
• pp.831-842
• /
• 1997

The purpose of the study was to test the reliability and validity of the Korean version of Task Self-Efficacy Scale for activities of dally living (ADL). The Task Self-Efficacy Scale was developed by Roberts(1996) for low-intensity exercise study with older people to predict their performance of ADL. The scale was translated and back translated by bilingual persons, and then was modified to resolve variations in the translations. The Korean version of Self-Efficacy Scale for ADL was then administered to 193 elderly people including 95 hospitalized patients and 98 outpatients or healthy people. Face to face interview was used to fill out the structured questionnaire, and each interview took approximately 30 minutes. The subjects for the study were 80 women and 112 men with an age range of 65 to 95 years(M=71 years) of whom 82.6% classified themselves as moderate or quite active Most subjects(80.2%) had an education level of elementary school or less. The Self-Efficacy Scale for ADL is measured on a 0 to 10 VAS, assessing three areas of ADL : self care activities, household tasks, and motor tasks. The higher the score is, the higher person's confidence in performing ADL. Psychometric testing revealed that the scale was found to be internally consistent, showing a Cronbach's alpha of .97 The scale was significantly correlated with subjects' level of activity and subjective assessment of their health status. Moderate correlation with health-related hardiness scale also supported the validity. Factor analysis was performed to confirm whether the scale represents the three sub-areas as suggested in the literature. The results of the factor analysis led to a three factor solution according to Kaiser's criterion, but the items were not strongly and cleanly loaded for the third factor. This can be explained in that, among the three sub-ADL areas of the self-efficacy scale, the areas of self care activities and household tasks seem to have similar levels of difficulty in performance with not enough differences for the self-efficacy scale to distinguish between the two areas. Therefore, one factor solution was suggested since ADL can be seen as a unit of activities at similar level of difficulty in performance. One factor solution explained 68.1% of variance of the 19-item scale and all items were correlated over .6 with the factor, showing that the selected factor solution fits the model. The results indicated that the Korean version of Task Self-Efficacy Scale for ADL was reliable and valid in producing useful information to evaluate the effects of various interventions toward promoting health and quality of life for elderly people.

• Research in Community and Public Health Nursing
• /
• v.19 no.2
• /
• pp.247-259
• /
• 2008

Purpose: This study was conducted to develop a teacher-efficacy scale which is adequate for heath education teachers. Method: A preliminary questionnaire was made based on the hypothetical factors. A total of 364 health education teachers answered to the preliminary test, and the potential factors of teacher-efficacy were checked out by exploratory factor analysis (EFA). In the main test 378 health education teachers answered, the factor structure was drawn by EFA, and confirmatory factor analysis (CFA) was conducted to test and the fitness of the factor model. Results: Through the preliminary test, 5 potential factors were found, which were instruction, health-service, administrative task, interpersonal relationship, and community connection. Also, 48 items were reduced to 31 items. Through the main test, from the 34 items were extracted 4 factors with 24 items, And then teacher-efficacy scale was developed, which included the subscales of instruction efficacy, health-service efficacy, task-interpersonal efficacy, and community connection efficacy. Conclusion: Different from that for other subject teachers, the factor structure for health teachers had 4 factors with 24 items. The scale developed in this study is consistent with health teachers' work areas, and for this reason, has significance as a adequate and valid scale to measure teacher-efficacy of health education teachers.

• 제어로봇시스템학회:학술대회논문집
• /
• 1996.10a
• /
• pp.1-7
• /
• 1996

This paper addresses the control problem of a mobile robot supporting a task robot with needs to be positioned precisely. The main difficulty residing in the precise control of a mobile robot supporting a task robot is providing an accurate and stable base for the task robot. That is, the end-plate of the mobile robot which is the base of the task robot can not be positioned accurately without external position sensors. This difficulty is resolved in this paper through the vision information obtained from the camera attached at the end of a task robot. First of all, the camera parameters were measured by using the images of a fixed object captured by the camera. The measured parameters include the rotation, the position, the scale factor, and the focal length of the camera. These parameters could be measured by using the features of each vertex point for a hexagonal object and by using the pin-hole model of a camera. Using the measured pose(position and orientation) of the camera and the given kinematics of the task robot, we calculate a pose of the end-plate of the mobile robot, which is used for the precise control of the mobile robot. Experimental results for the pose estimations are shown.

• Journal of Distribution Science
• /
• v.10 no.12
• /
• pp.19-24
• /
• 2012

Purpose - This paper discusses the heterogeneous fixed fleet vehicle routing problem with pick-up and delivery (HFFVRPPD), for vehicles with different capacities, fixed costs, and travel costs. Research Design, data, methodology - This paper made nine assumptions for establishing a mathematical model to describe HFFVRPPD. It established a practical mathematical model, and because of the non-deterministic polynomial-time hard (NP-hard), improved the traditional simulated annealing algorithm and tested a new algorithm using a certain scale model. Result - We calculated the minimum cost of the heterogeneous fixed fleet vehicle routing problem (HFFVRP) with a single task and, on comparing the results with the actual HFFVRP for the single task alone, observed that the total cost of HFFVRPPD reduced significantly by 46.7%. The results showed that the new algorithm provides better solutions and stability. Conclusions - This paper, by comparing the HFFVRP and HFFVRPPD results, highlights certain advantages of using HFFVRPPD in physical distribution enterprises, such as saving distribution vehicles, reducing logistics cost, and raising economic benefits.

• Journal of the Korea Institute of Building Construction
• /
• v.2 no.4
• /
• pp.153-161
• /
• 2002

Recently our government operates earned value management system (EVMS) for improving work management system of the construction projects over specific scale. The EVMS is based on the integrated management between construction cost and schedule. the existing systems, however, are focused on the cost management by using work quantity. This study suggests a work management model to integrate construction cost with activity information. The model introduces a work task concept as a tool that can connect construction rest to activity information. The suggested model in this study is verified by using actual data for the applicability to practical construction projects.