• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.517-521
• /
• 2023

최근 거대 언어모델의 등장과 동시에, 많은 매개변수를 효과적으로 학습하는 방법인 효율적인 매개변수 미세조정(Parameter Efficient Fine-Tuning) 연구가 활발히 진행되고 있다. 이 중에서 Adapter는 사전학습 언어모델(Pretrained Language Models)에 몇 개의 추가 병목 구조 모듈을 삽입하여 이를 학습하는 방식으로, 등장한 이후 다양한 연구 영역에서 주목받고 있다. 그러나 몇몇 연구에서는 병목 차원을 증가시켜 미세 조정보다 더 나은 성능을 얻는다는 주장이 나오면서, 원래의 의도와는 다른 방향으로 발전하고 있다는 의견도 있다. 이러한 맥락에서, 본 연구에서는 기존의 Adapter 구조를 개선한 MAdapter를 제안한다. MAdapter는 본래 Adapter에 중간 층을 추가하되 학습 가능한 매개변수의 수는 오히려 줄이는 방법으로, 전체 매개변수 수 대비 1% 내외 만을 학습에 활용하며, Adapter 대비 절반 정도의 매개변수만을 사용하여 기존 결과와 비슷하거나 더 나은 성능을 얻을 수 있는 것을 확인할 수 있다. 또한, 병목차원 크기 비교와 중간 층 개수 분석을 통한 최적의 MAdapter 구조를 찾고, 이로써 효율적인 매개변수 미세조정 방법을 제시한다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2018.05a
• /
• pp.40-40
• /
• 2018

현실화된 기후변화로 인하여 강수패턴 및 강수량의 변화가 발생함에 따라 물 관리의 불확실성이 증가하고 있는 실정이며 또한 2014년부터 시작된 극심한 가뭄은 안정적인 용수공급에 많은 어려움을 주고 있다. 예로 보령댐의 경우 저수율이 2015년 19%, 2017년 8.3%로 역대 최저를 기록한 바 있고 2018년 3월 기준 운문댐과 밀양댐의 경우 저수율이 각각 11%(심각), 26.2%(경계)로 극심한 물부족에 시달리고 있다. 이러한 극심한 가뭄에 대응하기 위해서는 시설간 연계 운영, 급수체계 조정, 도수로 가동 등과 같은 구조적 대책과 더불어 저수용량에 따라 용수공급량을 감축하는 형태의 비구조적 대책의 일환인 용수공급 조정기준이 동시에 이루어져야 한다. 용수공급 조정기준은 국토교통부에서 주기적으로 발생하고 있는 가뭄에 대하여 대응하기 위해 만들어진 기준으로 다목적댐의 저수량을 기준으로 단계별(관심, 주의, 경계, 심각) 공급량을 감축하여 장기간 공급함으로써 국가재난을 사전에 예방하는 것을 목표로 하고 있다. 이러한 용수공급 조정기준을 수립하기 위해서는 최소 20년 이상의 과거 관측된 유입량 자료를 모의 발생하여 이수안전도 95%이상을 만족하는 단계별 용수공급 조정 기준선을 작성하게 된다. 하지만 신규 댐의 경우 관측된 유입량 자료가 전무함에 따라 용수공급 조정 기준을 수립하는데 어려움이 있다. 따라서 본 연구에서는 신뢰할 수 있는 유입량 자료를 생산하기 위해 첫 번째로 물리적 기반인 분포형 강우-유출 모형을 구축하고 검보정을 수행하였으며 이처럼 구축된 강우-유출모형을 이용하여 다수의 유입량 자료를 생산하였다. 두 번째로 SAMS 2007을 이용하여 유입량 자료를 모의 발생시켰으며 이를 저수지 운영에 활용하여 이수안전도 95%를 만족하는 용수공급 조정기준을 수립하였다. 본 연구의 결과는 극심한 가뭄에 대비하여 안정적인 용수공급 및 수자원 확보 등 이수적인 측면에서 크게 활용 가능할 것으로 판단된다.

• The Journal of Bigdata
• /
• v.7 no.2
• /
• pp.11-29
• /
• 2022

Pre-trained language models are the most important and widely used tools in natural language processing tasks. Since those have been pre-trained for a large amount of corpus, high performance can be expected even with fine-tuning learning using a small number of data. Since the elements necessary for implementation, such as a pre-trained tokenizer and a deep learning model including pre-trained weights, are distributed together, the cost and period of natural language processing has been greatly reduced. Transformer variants are the most representative pre-trained language models that provide these advantages. Those are being actively used in other fields such as computer vision and audio applications. In order to make it easier for researchers to understand the pre-trained language model and apply it to natural language processing tasks, this paper describes the definition of the language model and the pre-learning language model, and discusses the development process of the pre-trained language model and especially representative Transformer variants.

• Management & Information Systems Review
• /
• v.30 no.4
• /
• pp.253-292
• /
• 2011

This study is significant of groping for the autonomous perspective of a medication system for the establishment of harmonious regional labor-management relationship and investigating a plan to minimize previously labor disputes through both analyzing mediation cases and finding out any core issue from the labor-management relationship, with matters experienced in the mediation field when I had served as a mediator. Specially, this study has also objectives to prevent previously any labor dispute through analyzing and minimizing the issue of the labor dispute by case and to establish harmonious labor-management relationship thereby. Further since it is necessary to understand the organization and roles of the regional labor relations commission in order to understand the concrete matters in respect of the labor dispute mediation, this study shall be to explore in detail the matters relating hereto and to acquire general knowledge of mediation through case analysis. Additionally, there is little studies exploring alternative research for the establishment of reasonable labor-management relationship with core issues referred a mediation to the labor relations commission through both the position heightening of the labor relations commission and the analysis of core issues until now. Thus, this study may provide a theoretical base for raising a technique to enhance negotiation skill through acquiring the previous training or full knowledge on the approaching manner to be taken by the labor and management in the collective bargaining or wage bargaining on the basis of items analyzed by core issue. The heightening perspectives to be acquired through the analysis of 50 or more mediation cases are as follows. First, it deems to be important what position each mediation party takes. Second, the information acquired by an investigator in the preliminary investigation before holding the mediation is very importantly utilized in the mediation. Third, the gumption of mediators in charge of the mediation is very helpful for the resolution of a case. Fourth, it shall be preceding to understand dispute issues. After reviewing fully the investigation report of an investigator, if separate review is required, it is tried to hold a separate meeting and then reduce the number of issues asserted by the labor & management and, if the number of such issues is reduced, the mediation may approach to be concluded. Fifth, it shall be kept in mind that any matter other than the scope of the law be based on not the judgement of mediator but the legal interpretation. Sixth, it is necessary for both labor and management parties to take a positive approach so as to make the healthy labor & management relationship anchored. Seventh, notwithstanding the mediators are part-time and take a service attitude, it shall be encouraged to abstain from slandering or inveighing against the mediators because the mediation is taken against oneself. Eighth, it is necessary to convert the management's recognition about a labor union. Ninth, it is necessary to not raise any issue on the matters, such as time-off system and multiple labor union, etc., which are legally enforced. Tenth, it is confirmed that the regional labor relations commission plays a bridge role of narrowing down the issue difference between the labor and management.

• Smart Media Journal
• /
• v.12 no.7
• /
• pp.76-83
• /
• 2023

This study proposes a method to assist webtoon artists in the process of webtoon creation by utilizing a pretrained Text-to-Image model to generate webtoon images from text. The proposed approach involves fine-tuning a pretrained Stable Diffusion model using a webtoon dataset transformed into the desired webtoon style. The fine-tuning process, using LoRA technique, completes in a quick training time of approximately 4.5 hours with 30,000 steps. The generated images exhibit the representation of shapes and backgrounds based on the input text, resulting in the creation of webtoon-like images. Furthermore, the quantitative evaluation using the Inception score shows that the proposed method outperforms DCGAN-based Text-to-Image models. If webtoon artists adopt the proposed Text-to-Image model for webtoon creation, it is expected to significantly reduce the time required for the creative process.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.3
• /
• pp.43-54
• /
• 2024

Recently, there has been a lot of research on utilizing Language Models, and Large Language Models have achieved innovative results in various tasks. However, the practical application faces limitations due to the constrained resources and costs required to utilize Large Language Models. Consequently, there has been recent attention towards methods to effectively utilize models within given resources. Curriculum Learning, a methodology that categorizes training data according to difficulty and learns sequentially, has been attracting attention, but it has the limitation that the method of measuring difficulty is complex or not universal. Therefore, in this study, we propose a methodology based on data heterogeneity-based Curriculum Learning that measures the difficulty of data using reliable prior information and facilitates easy utilization across various tasks. To evaluate the performance of the proposed methodology, experiments were conducted using 5,000 specialized documents in the field of information communication technology and 4,917 documents in the field of healthcare. The results confirm that the proposed methodology outperforms traditional fine-tuning in terms of classification accuracy in both LoRA fine-tuning and full fine-tuning.

• The Korean Society of Law and Medicine
• /
• v.16 no.1
• /
• pp.255-288
• /
• 2015

It is desirable to prevent medical accidents because they bring about irretrievable outcomes to patients, as they are directly related to each patient's life, and health. However, once medical accidents occur, it is appropriate to resolve them quickly without conflict before the feelings of directly involved people are intensely confronted with each other. Korea Consumer Agency carries out medical dispute mediation to address such disputes quickly, fairly, and efficiently, and so does Korea Medical Dispute Mediation and Arbitration Agency. Although there has been constant debate on a merge between the two agencies because of duplicated work and consequent inefficiency, it is desirable to maintain the two agencies to ensure consumers' options and to promote the mutual development of the agencies through competition. Therefore, there should be legal and systematical support for Korea Consumer Agency to have fair competition with Korea Medical Dispute Mediation and Arbitration Agency. This is not for Korea Consumer Agency, but ultimately for consumers.

• Journal of IKEEE
• /
• v.19 no.1
• /
• pp.33-40
• /
• 2015

Inverse problem in electrical impedance tomography (EIT) is highly ill-posed therefore prior information is used to mitigate the ill-posedness. Regularization methods are often adopted in solving EIT inverse problem to have satisfactory reconstruction performance. In solving the EIT inverse problem, iterative Gauss-Newton method is generally used due to its accuracy and fast convergence. However, its performance is still suboptimal and mainly depends on the selection of regularization parameter. Although, there are few methods available to determine the regularization parameter such as L-curve method they are sometimes not applicable for all cases. Moreover, regularization parameter is a scalar and it is fixed during iteration process. Therefore, in this paper, a novel method is used to determine the regularization parameter to improve reconstruction performance. Conductivity norm is calculated at each iteration step and it used to obtain the regularization parameter which is a diagonal matrix in this case. The proposed method is applied to human thorax imaging and the reconstruction performance is compared with traditional methods. From numerical results, improved performance of proposed method is seen as compared to conventional methods.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.9 no.2
• /
• pp.99-105
• /
• 1989

This paper shows the possibility to have a accurate record and analysis of all data of pre/post-environmental impact assessment. For the experimental test, a surveying was carried with aerial photos by non-metric camera of remote controlled model airplene, aeial camera, LANDSAT data. The three approaches are briefly described. Finally, author suggest an environmental impact assessment method using aerial photogrammetric method. This will lead to more accurate and effective method in the environmental impact assessment.

• Proceedings of the Korean Society of Disaster Information Conference
• /
• 2016.11a
• /
• pp.275-277
• /
• 2016

본 논문에서는 Door Fan Test 수행 사례를 중심으로 사전 준비에 필요한 사항에 대해 분석하였다. 분석 결과 관통부 주위 개구부, 벽체 틈새 개구부 등 방화구역의 누설부위를 통해 약제가 누출되는 문제가 있었다. 원인으로는 전기배선 공사로 인해 내화충전된 부분을 제거하고 공사 후 다시 충전하지 않는 경우, 벽과 천장사이의 틈을 내화 실란트 등으로 마감을 하지 않은 경우가 대부분이었다. 이와 같이 누설부위가 많은 경우 테스트의 결과 불합격이 나오게 되며, 이러한 문제점을 개선하기 위해 사전현장조사를 통해 방호구역의 누설부위를 조사하고 보완조치를 진행한 후 테스트를 수행하는 절차가 필요하였다. 한편, 테스트 운영에 관한 사항으로는 해당 실(room)을 위한 충분한 유량 미확보, 장애물로 인한 와류형성, 차압측정구의 설치문제 등이 있었다. 이를 개선하기 위해서 공기를 유입하는 실은 외부로부터 충분한 공기유입이 있어야하며, 팬 앞의 장애물을 제거하거나 팬의 설치 위치를 조정하는 것이 필요하며, 공기의 유동이 가장 적은 곳에 측정구를 설치하는 등의 조치가 필요하다.