• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.162-167
• /
• 2023

본 논문은 거대언어모델에 대한 한국어 상식추론 기반의 새로운 평가 방식을 제안한다. 제안하는 평가 방식은 한국어의 일반 상식을 기초로 삼으며, 이는 거대언어모델이 주어진 정보를 얼마나 잘 이해하고, 그에 부합하는 결과물을 생성할 수 있는지를 판단하기 위함이다. 기존의 한국어 상식추론 능력 평가로 사용하던 Korean-CommonGEN에서 언어 모델은 이미 높은 수준의 성능을 보이며, GPT-3와 같은 거대언어모델은 사람의 상한선을 넘어선 성능을 기록한다. 따라서, 기존의 평가 방식으로는 거대언어모델의 발전된 상식추론 능력을 정교하게 평가하기 어렵다. 더 나아가, 상식 추론 능력을 평가하는 과정에서 사회적 편견이나 환각 현상을 충분히 고려하지 못하고 있다. 본 연구의 평가 방법은 거대언어모델이 야기하는 문제점을 반영하여, 다가오는 거대언어모델 시대에 한국어 자연어 처리 연구가 지속적으로 발전할 수 있도록 하는 상식추론 벤치마크 구성 방식을 새롭게 제시한다.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.20 no.2
• /
• pp.205-213
• /
• 2018

Development and validation of crop models often require measurements of biomass for the crop of interest. Considerable efforts would be needed to obtain a reasonable amount of biomass data because the destructive sampling of a given crop is usually used. The Kinect sensor, which has a combination of image and depth sensors, can be used for estimating crop biomass without using destructive sampling approach. This approach could provide more data sets for model development and validation. The objective of this study was to examine the applicability of the Kinect sensor for estimation of chinese cabbage fresh weight. The fresh weight of five chinese cabbage was measured and compared with estimates using the Kinect sensor. The estimates were obtained by scanning individual chinese cabbage to create point cloud, removing noise, and building a three dimensional model with a set of free software. It was found that the 3D model created using the Kinect sensor explained about 98.7% of variation in fresh weight of chinese cabbage. Furthermore, the correlation coefficient between estimates and measurements were highly significant, which suggested that the Kinect sensor would be applicable to estimation of fresh weight for chinese cabbage. Our results demonstrated that a depth sensor allows for a non-destructive sampling approach, which enables to collect observation data for crop fresh weight over time. This would help development and validation of a crop model using a large number of reliable data sets, which merits further studies on application of various depth sensors to crop dry weight measurements.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.6
• /
• pp.41-50
• /
• 2003

For copyright protection of digital contents, we employ watermarking techniques to embed watermark signals into digital host data. In this paper we propose an adaptive watermarking algorithm for three-dimensional (3-D) mesh models. Watermark signals are inserted into vertex coordinates adaptively according to changes of their position values. While we embed strong watermarks in the areas of large variations, watermarks are weakly inserted in other areas. After we generate triangle strips by traversing the 3-D model and convert the Cartesian coordinates to the spherical coordinate system, we calculate variations of vertex positions along the traversed strips. Then, we insert watermark signals into the vertex coordinates adaptively according to the calculated variations. We demonstrate that imperceptibility of the inserted watermark is significantly improved and show the bit error rate (BER) for robustness.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.16 no.2
• /
• pp.179-184
• /
• 2006

An intuitive understanding of the dynamic pattern generation in asymmetric networks may be considered an essential component in developing models for the dynamic information processing. It has been reported that the neural network with cyclic connections generates multiple limit cycles. The dynamics of discrete time network with cyclic connections has been investigated intensively. However, the dynamics of a cyclic connection neural network in continuous-time has not been well-known due to the considerable complexity involved in its calculation. In this paper, the dynamic behavior of a continuous-time cyclic connection neural network, in which each neuron is connected only to its nearest neurons with binary synaptic weights of ${\pm}1$, has been investigated. Furthermore, the dynamics and stability of the network have been analyzed using a piece-wise linear approximation.

• The KIPS Transactions:PartD
• /
• v.11D no.3
• /
• pp.671-682
• /
• 2004

As new and diverse information technologies are being introduced and software complexity is increased, software development cost and efforts are also sharply increased. Component-Based Development (CBD) technology is appealing as a new way to reduce the cost and effort by increasing reusability and maintainability. Component in CBD has variability infernally which enables customization of the component within the specific domain. A component user can easily set up internally variability parts though Required interface which is provided by the component. Enterprise JavaBeans (EJB) is utilized as a commercial standard to implement Java-based components. However, EJB constructs are limited in directly implementing Required interfaces of coarse-grained components［8］. In this paper, we define Required interface and propose interface-storage technique, class-storage technique, interface-generation technique, and Plug -in technique for implementing required interface of component. Interface-storage technique stores variable value in Required interface and class-storage technique take the Bean containg variability as Required interface without modification of component model. Interface-generation technique generates new Bean which takes the role of Required interface for component variability and Plug-in technique sets up component variability that component user plugged-in variable part externally The proposed four techniques conform to the semantics of CBD component interface and enable the implementation of high quality components.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.41 no.6
• /
• pp.1-10
• /
• 2004

A haptic rendering system is proposed for a puncture task of a virtual vertebra model. To build a mesh model from medical images, Delaunay triangulation is applied and physical models are based on elasticity theory. Also, a redundant actuated 6 DOF parallel type haptic device is designed to display large force and to resolve the singularity problem of parallel type mechanisms. Haptic feeling of puncture task and the performance of the proposed haptic device are tested by two puncture task experiments.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.373-373
• /
• 2022

전세계적으로 기후변화로 인해 3년 이상의 기간동안 지속되는 다년 가뭄의 빈도와 심도가 증가하고 있으며, 이로 인한 피해도 증가하고 있다. 본 연구에서는 이를 반영하여 전국 다목적댐 및 용수댐에서 모두 주요 가뭄 대응 대책으로 사용되고 있는 현행 용수공급 조정기준을 개선하는 방안을 제안하고자 한다. 가장 먼저, 장기 기억 반영이 가능한 시계열 모형인 ARFIMA(Autoregressive Fractional Integrated Moving Average) 모델을 사용하여 다양한 강도의 장기 기억을 가지고 있는 연간 유입량을 생성하였다. 이후, 연간 유입량을 k-최근접 이웃 방법 기반의 배분 도구를 사용하여 10일 단위 유입량으로 분배하였으며 이를 대체 용수공급 조정기준을 생성하기 위한 입력 변수로 사용하였다. 새로운 용수공급 조정기준은 매 시점마다 새롭게 업데이트되는 정보를 통해 현행 기준과 함께 적응형으로 저수지 운영에 사용되었다. 다년 가뭄이 반영된 유입량으로 적응형으로 저수지 운영을 관측 유입량 하에서 빈도와 크기의 측면에서 분석을 시행하였다. 그 결과, 심각한 실패(물 부족 비율 30% 이상)의 빈도의 경우 현행 기준 운영 시 6.14%에서 적응형 운영 시행 시 2.99%로 개선되었지만, 전체 기간 동안의 신뢰도는 적응형 운영보다(26.42%) 현행 운영 하에서 더욱 나은 결과를 보였다(41.19%). 위와 같은 분석 결과는 심각한 실패의 빈도와 크기를 줄이는 용수공급 조정기준을 시행하는 원론적인 목적과 일치하기에, 본 연구에서 제안하는 다년 가뭄에 대비한 적응형 운영 방안은 향후 길게 지속되는 가뭄 조건에서 저수지 운영 정책으로 활용될 수 있음을 확인하였다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.6B
• /
• pp.1098-1107
• /
• 1999

In multimedia applications, a source usually generates multiple streams. By heterogeneous multimedia multicast, we mean a recipient can receive some of them, not necessarily all of them. A recipient bids for what it wants to receive and the source gains the same amount when a connection is established. The problem of distributing streams for heterogeneous multicast to maximize the source's gain, can be solved using a 0-1 integer programming, hewn as NP-complete. In this paper, we propose efficient stream distribution algorithms in two different types of multicast models. The first restricted model assumes that the capacity for a link in the multicast tree is grater than or equal to the capacities of its descendant links. In the second unrestricted model, we drop out the restriction in the restricted model. Proposed algorithms have better time and space complexities compared with any existing one. In addition, distributed implementations are straightforward, which is very useful for large networks.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.5
• /
• pp.827-832
• /
• 2024

The AI program Sora is a video production model that can be used innovatively and is the starting point of a major paradigm shift in video planning and production in the future. In this paper, through consideration of the characteristics, application, and process of the AI video production program, the characteristics of the AI design video production method were understood, and the production algorithm was considered. The detailed consideration and characteristics of the work creation process for the video graphic AI video generation program that will be intensified every year were examined. Next, the method of generating a customized video with a text prompt and the process of innovative production results different from the previous production method were considered. In addition, the design direction through the generation of AI images was studied through the review of the strengths and weaknesses of the image details of the recently announced AI music video results. By considering the security of the AI generation video Sora and looking at the internal process of the actual AI process, it will be possible to present indicators for the future direction of AI video model production and education along with the direction of the design designer and education system. In the text and conclusion, we analyzed the strengths and weaknesses and future status of OpenAI Sora image, concluded how to apply the Sora model's capabilities, limitations, quality, and human creativity, and presented problems and alternatives through examples of the Sora model's capabilities and limitations to increase human creativity.

• Journal of Radiation Protection and Research
• /
• v.11 no.2
• /
• pp.139-145
• /
• 1986

An improved mothod of assessing fuel status by analyzsis of the fission product in the reactor coolant system is proposed. The release mechanism of specific fission products is established for determination of the coefficients in the equations which relate the radioactivities with the amount of defected fuel. Knock-out and migration models are employed in the formulation of the release mechanism. The influence of the tramp uranium is quantified. Sample calculations were made for KNU 1 reactor system using the I-131 and I-133 concentrations in the primary coolant. The estimated number of defected fuel pins in the third and sixth cycles appeared to be $9.34{\pm}1.13\;and\;0.294{\pm}0.092$, respectively.