• 한국컴퓨터정보학회논문지
• /
• 제29권3호
• /
• pp.43-54
• /
• 2024

최근 언어모델을 활용하기 위한 연구가 활발히 이루어지며, 큰 규모의 언어모델이 다양한 과제에서 혁신적인 성과를 달성하고 있다. 하지만 실제 현장은 거대 언어모델 활용에 필요한 자원과 비용이 한정적이라는 한계를 접하면서, 최근에는 주어진 자원 내에서 모델을 효과적으로 활용할 수 있는 방법에 주목하고 있다. 대표적으로 학습 데이터를 난이도에 따라 구분한 뒤 순차적으로 학습하는 방법론인 커리큘럼 러닝이 주목받고 있지만, 난이도를 측정하는 방법이 복잡하거나 범용적이지 않다는 한계를 지닌다. 따라서, 본 연구에서는 신뢰할 수 있는 사전 정보를 통해 데이터의 학습 난이도를 측정하고, 이를 다양한 과제에 쉽게 활용할 수 있는 데이터 이질성 기반 커리큘럼 러닝 방법론을 제안한다. 제안방법론의 성능 평가를 위해 국가 R&D 과제 전문 문서 중 정보통신 분야 전문 문서 5,000건, 보건의료전문 문서 데이터 4,917건을 적용하여 실험을 수행한 결과, 제안 방법론이 LoRA 미세조정과 전체 미세조정 모두에서 전통적인 미세조정에 비해 분류 정확도 측면에서 우수한 성능을 나타냄을 확인했다.

• Journal of Communications and Networks
• /
• 제11권2호
• /
• pp.134-147
• /
• 2009

Tracking and exploiting instantaneous spectrum opportunities are fundamental challenges in opportunistic spectrum access (OSA) in presence of the bursty traffic of primary users and the limited spectrum sensing capability of secondary users. In order to take advantage of the history of spectrum sensing and access decisions, a sequential decision framework is widely used to design optimal policies. However, many existing schemes, based on a partially observed Markov decision process (POMDP) framework, reveal that optimal policies are non-stationary in nature which renders them difficult to calculate and implement. Therefore, this work pursues stationary OSA policies, which are thereby efficient yet low-complexity, while still incorporating many practical factors, such as spectrum sensing errors and a priori unknown statistical spectrum knowledge. First, with an approximation on channel evolution, OSA is formulated in a multi-armed bandit (MAB) framework. As a result, the optimal policy is specified by the wellknown Gittins index rule, where the channel with the largest Gittins index is always selected. Then, closed-form formulas are derived for the Gittins indices with tunable approximation, and the design of a reinforcement learning algorithm is presented for calculating the Gittins indices, depending on whether the Markovian channel parameters are available a priori or not. Finally, the superiority of the scheme is presented via extensive experiments compared to other existing schemes in terms of the quality of policies and optimality.

• 전기전자학회논문지
• /
• 제23권1호
• /
• pp.235-241
• /
• 2019

본 논문에서는 농업 기계 시스템에서 사용하기 위한 딥러닝 알고리즘 기반의 프레임 내의 관심 영역 유사성을 이용한 새로운 후방 접근 차량 검출 알고리즘을 제안한다. 농업 기계 시스템은 후방에서 접근하는 차량만 검출해야 한다. 지나가는 자동차가 검출되면 혼란을 야기할 수 있다. 논문에서는 차량 검출을 위해 딥러닝에서 뛰어난 검출률을 나타내는 Faster R-CNN 모델을 사용하였다. 딥러닝은 뒤에서 접근하는 차량뿐만 아니라 지나가는 차량도 검출하므로 긍정오류 차량을 배제해야 한다. 본 논문에서 이를 해결하기 위해 검출된 프레임에서 관심 영역에 대한 유사성과 평균 에러를 피라미드 형태로 이용하여 접근하는 자동차만 검출하는 알고리즘을 제안하였다. 실험을 통하여 제안된 방법이 평균 98.8%의 높은 검출률을 나타내었다.

• Structural Engineering and Mechanics
• /
• 제83권4호
• /
• pp.537-549
• /
• 2022

In this paper, a novel multi-feature model predictive control (MPC) framework with real-time and adaptive performances is proposed for intelligent structural control in which some drawbacks of the algorithm including, complex control rule and non-optimality, are alleviated. Hence, Linear Programming (LP) is utilized to simplify the resulted control rule. Afterward, the Whale Optimization Algorithm (WOA) is applied to the optimal and adaptive tuning of the LP weights independently at each time step. The stochastic control rule is also achieved using Kalman Filter (KF) to handle noisy measurements. The Extreme Learning Machine (ELM) is then adopted to develop a data-driven and real-time control algorithm. The efficiency of the developed algorithm is then demonstrated by numerical simulation of a twenty-story high-rise benchmark building subjected to earthquake excitations. The competency of the proposed method is proven from the aspects of optimality, stochasticity, and adaptivity compared to the KF-based MPC (KMPC) and constrained MPC (CMPC) algorithms in vibration suppression of building structures. The average value for performance indices in the near-field and far-field (El earthquakes demonstrates a reduction up to 38.3% and 32.5% compared with KMPC and CMPC, respectively.

• Asia pacific journal of information systems
• /
• 제33권4호
• /
• pp.1171-1187
• /
• 2023

This study explores how an organization has successfully implemented ML-supported HR services to resolve high employee turnover problems in the IT sector. The empirical setting of the research is where contradicting institutional logics exist among technical, HR, and business groups regarding the ML model development and use of the model predictions in HR services. Institutional framework is used to identify the roles of organizational actors and the legitimacy structures in the organizational environments that can shape or constrain the ML led organizational changes. In institutional theories, technology adoption and organizational change are not only constrained by organizational context, but also fostered through organizational actors' roles and efforts to increase the legitimacy for the change. This research found that when multiple contradicting institutional logics exist, legitimizing the establishment of an enabling environment for multiple logics to reconcile and for the project to move forward is critical. Industry-wide conditions, previous experiences with the pilot ML project, forming a TFT with clearly defined roles and responsibilities, and relevant KPIs are found to legitimize the HR team and the business division to collaborate with the technical personnel to launch ML-supported HR services.

• Journal of Electrical Engineering and Technology
• /
• 제1권2호
• /
• pp.200-210
• /
• 2006

This paper proposes the motion generation method in which the movement of the 3-links leg subsystem in constrained to slider-link and a singular posture can be easily avoided. This method is the realization of jumping control moving in a vertical direction, which mimics a cat's behavior. To consider the movement from the point of the constraint mechanical system, a robotics system for realizing the motion will change its configuration according to the position. The effectiveness of the proposed scheme is illustrated by simulation and experimental results.

• 대한음성학회지:말소리
• /
• 제66호
• /
• pp.73-86
• /
• 2008

Beamforming is one of the spatial filtering techniques which extract only desired signals from noisy environments using microphone arrays. Fixed beamforming is a simple concept and easy to implement. However, it does not show good performance in real noisy conditions. As an adaptive beamforming, Frost algorithm can be a good candidate. It uses the concept of the linearly constrained minimum variance (LCMV) algorithm. The difference between the Frost and the LCMV algorithm is the error correction scheme which is very effective feature in the aspect of performance. In this paper, as quadrature mirror filtering (QMF)-based filterbank is utilized as the pre-processing of the Frost beamformning, the filter length and the learning rate of each band is optimized to improve the performance. The performance is measured by the signal-to-noise ratio (SNR) and the Bark's scale spectral distortion (BSD).

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2004년도 가을 학술발표논문집 Vol.31 No.2 (1)
• /
• pp.112-114
• /
• 2004

지능형 로봇 에이전트 기술이 발전하면서 서비스 질을 높이기 위한 방법으로 컨텍스트의 활용성이 부각되고 있다. 하지만 컨텍스트 분류 기술들은 아직까지 초기 개발 단계이며 다양한 방법들이 시도되고 있다. 본 논문에서는 전문가의 지식과 학습된 지식을 함께 적용할 수 있고 사람이 그 내용을 이해하기 유리한 베이지안 네트워크(BN)를 이용한 컨텍스트 분류 방법을 제안한다. 일반적인 BN 구조 학습에 사전 지식 및 방향성, 연결 관계 범위를 부여할 수 있는 제한(Constraint)을 적용한 효율적인 컨텍스트 분류 방법을 소개하고, 몇 가지 비교 실험을 통해 기존 방법에 비해 전문가의 개입이 줄어들고 좀 더 신뢰성 있는 컨텍스트 분류기를 얻을 수 있음을 보인다.

• Advances in Computational Design
• /
• 제4권3호
• /
• pp.197-210
• /
• 2019

Motion generation of a four-bar linkage is a type of mechanism synthesis that has a wide range of applications such as a pick-and-place operation in manufacturing. In this research, the use of meta-heuristics for motion generation of a four-bar linkage is demonstrated. Three problems of motion generation were posed as a constrained optimization probably using the weighted sum technique to handle two types of tracking errors. A simple penalty function technique was used to deal with design constraints while three meta-heuristics including differential evolution (DE), self-adaptive differential evolution (JADE) and teaching learning based optimization (TLBO) were employed to solve the problems. Comparative results and the effect of the constraint handling technique are illustrated and discussed.

• Nuclear Engineering and Technology
• /
• 제55권2호
• /
• pp.443-451
• /
• 2023

Machine Learning has introduced many solutions in data science, but its application in IoT faces significant challenges, due to the limitations in memory size and processing capability of constrained devices. In this paper we design an automatic gamma radiation detection and identification embedded system that exploits the power of TinyML in a SiPM micro radiation sensor leveraging the Edge Impulse platform. The model is trained using real gamma source data enhanced by software augmentation algorithms. Tests show high accuracy in real time processing. This design has promising applications in general-purpose radiation detection and identification, nuclear safety, medical diagnosis and it is also amenable for deployment in small satellites.