• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.408-410
• /
• 2021

In the NLP field, the pre-training model BERT launched by the Google team in 2018 has shown amazing results in various tasks in the NLP field. Subsequently, many variant models have been derived based on the original BERT, such as RoBERTa, ERNIEBERT and so on. In this paper, the WWMBERT (Whole Word Masking BERT) model suitable for Chinese text tasks was used as the baseline model of our experiment. The experiment is mainly for "Text-level Chinese text classification tasks" are improved, which mainly combines Tapt (Task-Adaptive Pretraining) and "Multi-Sample Dropout method" to improve the model, and compare the experimental results, experimental data sets and model scoring standards Both are consistent with the official WWMBERT model using Accuracy as the scoring standard. The official WWMBERT model uses the maximum and average values of multiple experimental results as the experimental scores. The development set was 97.70% (97.50%) on the "text-level Chinese text classification task". and 97.70% (97.50%) of the test set. After comparing the results of the experiments in this paper, the development set increased by 0.35% (0.5%) and the test set increased by 0.31% (0.48%). The original baseline model has been significantly improved.

• Journal of Navigation and Port Research
• /
• v.36 no.5
• /
• pp.363-370
• /
• 2012

Recently, International Maritime Organization has been developing e-navigation implementation strategy plan, which is focused on various services for vessel safety navigation. Then, different kinds of software will be developed in maritime area and with this, the quality issues are to be expected becoming more important. In this paper, we adopt software refactoring techniques to reduce the complexity of structure on source code level. It makes software program more effective to understand and modify, without any change of outward behavior. The existing AIS broadcast server program is used as an example for our trial, and calculating coupling and cohesion metric are introduced to analyze the refactoring effect, taking account of the maintainability of IEC/ISO9126 software quality standards.

• Journal of electromagnetic engineering and science
• /
• v.13 no.4
• /
• pp.251-258
• /
• 2013

A procedure for antenna parameter selections is proposed that considers the relationships between synthetic aperture radar performance and the antenna parameters of a parabola-type reflector antenna with a central flat dish. The effects of a central dish designed for weight reduction on the antenna beam pattern are also quantitatively analyzed using commercially available software based on the physical optics algorithm. The results of the theoretical analysis and simulation predict that a larger size of the central dish results in an increase in the sidelobe level, which is the reason for the increase in two important ambiguities, such as range ambiguity ratio (RAR) and azimuth ambiguity ratio (AAR). The dependence of RAR and AAR on Pulse repetition frequency is also analyzed and discussed.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.9
• /
• pp.96-105
• /
• 2015

This paper proposes a DSP optimization solution of rain detection and removal algorithm. We propose rain detection and removal algorithms considering camera motion, and also presents optimization results in algorithm level and DSP level. At algorithm level, this paper utilizes a block level binary pattern analysis, and reduces the operation time by using the fast motion estimation algorithm. Also, the algorithm is optimized at DSP level through inter memory optimization, EDMA, and software pipelining for real-time operation. Experiment results show that the proposed algorithm is superior to the other algorithms in terms of visual quality as well as processing speed.

• Geomechanics and Engineering
• /
• v.4 no.1
• /
• pp.1-18
• /
• 2012

The spatial variation of ground motion in Kolkata Metropolitan District (KMD) has been estimated by generating synthetic ground motion considering the point source model coupled with site response analysis. The most vulnerable source was identified from regional seismotectonic map for an area of about 350 km radius around Kolkata. The rock level acceleration time histories at 121 borehole locations in Kolkata for the vulnerable source, Eocene Hinge Zone, due to maximum credible earthquake (MCE) moment magnitude 6.2 were generated by synthetic ground motion model. Soil investigation data of 121 boreholes were collected from the report of Soil Data Bank Project, Jadavpur University, Kolkata. Surface level ground motion parameters were determined using SHAKE2000 software. The results are presented in the form of peak ground acceleration (PGA) at rock level and ground surface, amplification factor, and the response spectra at the ground surface for frequency 1.5 Hz, 3 Hz, 5 Hz and 10 Hz and 5% damping ratio. Site response study shows higher PGA in comparison with rock level acceleration. Maximum amplification in some portion in KMD area is found to be as high as 3.0 times compared to rock level.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.7
• /
• pp.353-360
• /
• 2017

In this study, a software education program based on blended learning using scratch was designed for pre-service elementary teachers' software education. Software education program was applied to experimental group and control group within the university of education using Scratch programming for 12 weeks. Blended learning using online lectures was applied to experimental group. The pre-service teachers' recognition about software education and self-evaluation of computational thinking were performed. Then, we analyzed the effect of the developed education program on the recognition of software education and computational thinking. As a result, the level of post-recognition of software education in the experimental group was significantly higher than that of the pre-recognition. In the self-evaluation of computational thinking, the experimental group was significantly higher than the control group. Therefore, it can be seen that the software education program based on blended learning can help improve the learner's computational thinking.

• Journal of Digital Convergence
• /
• v.8 no.1
• /
• pp.93-107
• /
• 2010

Performance of software development is measured by two major criteria - roductivity and quality. Although the criteria is empirically tested in software engineering research, they often present with a limited way under consideration of a few factors or contexts for developers to focus on the either productivity facets or quality facets. Analyzing data on software development performance collected over a 13-year period from 20 countries, we investigated how major software development factors - development type, development platform, development technique, language type, DBMS, methodology, methodology acquisition, CASE,, summary of work effort, resource level, max team size, affect the performance of software development. The results suggest that productivity and quality of software development are affected by different factors and context: function points, line of code, extreme defects, major defects, or minor defects. This research provides the empirical evidence that the two performance criteria require for software developer to have careful attention to find the optimal balance between the two performance criteria.

• Aerospace Engineering and Technology
• /
• v.7 no.1
• /
• pp.83-90
• /
• 2008

The next generation LID satellite has 64KB PROM which contains the boot loader and the monitor software, and two 4MB NVMEMs which are used for flight software storage. The boot loader has two operation modes which are the flight software mode and the monitor mode. In the flight software mode, it checks CRC checksum of selected NVMEM and copies flight software image from NVMEM to RAM And then it starts VxWorks RTOS in RAM, creates flight software tasks, and starts execution of flight software. In the monitor mode, it activates monitor software which performs NVMEM reprogramming and board-level testing on the ground. This paper is to present the design of Boot ROM software and verification method using simulator.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.9
• /
• pp.1240-1248
• /
• 2013

This paper describes a switching-level operation analysis of BTB(Back-To-Back) converter for HVDC(high voltage DC) application based on MMC(modular multi-level converter). A switching-level operation analysis for BTB converter is very important to understand the converter operation in detail and check the voltage and current transients in each components. However, the development of switching-level simulation model for the actual size BTB Converter is very difficult because the MMC normally has more than 150 sub-modules for each arm. So, a switching level simulation model for the 11-level MMC-based BTB converter was developed with PSCAD/EMTDC software, which has 12 sub-modules for the positive arm and another 12 sub-modules for the negative arm. The DC-voltage balance algorithm, the circulating-current reduction algorithm, the harmonic reduction algorithm, and the redundancy operation algorithm were included in this simulation model. The developed simulation model can be utilized to analyze the MMC-based BTB converter for HVDC application in switching level and to develop the protection scheme for the MMC-based BTB converter for HVDC application.

• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1503-1521
• /
• 2015

Agent-based modeling and simulation (ABMS) is used to model of market and social phenomena by utilizing agents' fine-grained behaviors and interactions that cannot be implemented in a conventional simulation. However, ABMS represents irrational agents and hinders the achievement of individual or overall goals since ABMS is based on agent-based software, which follows the principle of rationality at the knowledge level [1]. This problem was solved in the agent-based software engineering (ABSE) field by using behavior laws for the social level [2]. However, they still do not propose the specific development methodology for how to develop the social level in a systematic way. Therefore, in order to propose agent-based modeling and simulation methods that reflect the behavior laws of social level characteristics, our study used the Tropos that can combine ABSE and social behavior laws for the presentation of concrete tasks and deliverables for each development step by step. In addition, the proposed method will be specified through experiments with specific application examples and case studies on the self-adaptive smart grid and the military domain system.