• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.387-390
• /
• 1998

In this paper, we suggested a neuro-fuzzy learning algorithm for tuning fuzzy rules, in which a fuzzy system model is of additive-type. Using the method, it is possible to reduce the computation size, since performing the fuzzy inference and tuning the fuzzy rules of each fuzzy subsystem model are independent. Moreover, the efficiency of suggested method is shown by means of a numerical example.

• Phonetics and Speech Sciences
• /
• v.15 no.3
• /
• pp.83-88
• /
• 2023

This study investigates the fine-tuning of large-scale Automatic Speech Recognition (ASR) models, specifically OpenAI's Whisper model, for domain-specific applications using the KsponSpeech dataset. The primary research questions address the effectiveness of targeted lexical item emphasis during fine-tuning, its impact on domain-specific performance, and whether the fine-tuned model can maintain generalization capabilities across different languages and environments. Experiments were conducted using two fine-tuning datasets: Set A, a small subset emphasizing specific lexical items, and Set B, consisting of the entire KsponSpeech dataset. Results showed that fine-tuning with targeted lexical items increased recognition accuracy and improved domain-specific performance, with generalization capabilities maintained when fine-tuned with a smaller dataset. For noisier environments, a trade-off between specificity and generalization capabilities was observed. This study highlights the potential of fine-tuning using minimal domain-specific data to achieve satisfactory results, emphasizing the importance of balancing specialization and generalization for ASR models. Future research could explore different fine-tuning strategies and novel technologies such as prompting to further enhance large-scale ASR models' domain-specific performance.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.289-294
• /
• 2023

대형 언어 모델(Large Language Model)은 방대한 데이터와 파라미터를 기반으로 문맥 이해에서 높은 성능을 달성하였지만, Human Alignment를 위한 문장 생성 제어 연구는 아직 활발한 도전 과제로 남아있다. 본 논문에서는 Instruction Tuning을 통한 문장 생성 제어 실험을 진행한다. 자연어 처리 도구를 사용하여 단일 혹은 다중 제약 조건을 포함하는 Instruction 데이터 셋을 자동으로 구축하고 한국어 언어 모델인 Polyglot-Ko 모델에 fine-tuning 하여 모델 생성이 제약 조건을 만족하는지 검증하였다. 실험 결과 4개의 제약 조건에 대해 평균 0.88의 accuracy를 보이며 효과적인 문장 생성 제어가 가능함을 확인하였다.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.56 no.11
• /
• pp.2010-2016
• /
• 2007

PID control is windely used to control stable processes, However, its application to integrating processes is less common. In this paper, we proposed a stable PID controller tuning method for integrating processes with time delay using model reduction method. For proposed model reduction method, it disconnect an integrating factor from integrating processes and reduces separate process using reduction method. and it connect an integrating factor to reduced model. We can obtain stable integrating processes using P controller in inner feedback loop and PID tuning is then used to cancel the pole of the feedback loop. This guarantees both robustness and performance. Simulation examples are given to show the good performance of the proposed tuning method comparing with other methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.12
• /
• pp.1769-1776
• /
• 2022

Automatically summarizing long sentences is an important technique. The BART model is one of the widely used models in the summarization task. In general, in order to generate a summarization model of a specific domain, fine-tuning is performed by re-training a language model trained on a large dataset to fit the domain. The fine-tuning is usually done by changing the number of nodes in the last fully connected layer. However, in this paper, we propose a fine-tuning method by adding an attention layer, which has been recently applied to various models and shows good performance. In order to evaluate the performance of the proposed method, various experiments were conducted, such as accumulating layers deeper, fine-tuning without skip connections during the fine tuning process, and so on. As a result, the BART model using two attention layers with skip connection shows the best score.

• Journal of Institute of Control, Robotics and Systems
• /
• v.8 no.12
• /
• pp.1056-1060
• /
• 2002

Model-based tuning rules of the PID controller are proposed incorporating with real-coded genetic algorithms. The optimal parameter sets of the PID controller for step set-point tracking are obtained based on the first-order time delay model and a real-coded genetic algorithm as an optimization tool. As for assessing the performance of the controllers, performance indices(ISE, IAE and ITAE) are adopted. Then tuning rules are derived using the tuned parameter sets, potential rule models and another real-coded genetic algorithm A set of simulation works is carried out to verify the effectiveness of the proposed rules.

• Proceedings of the Korean Society of Marine Engineers Conference
• /
• 2002.05a
• /
• pp.261-266
• /
• 2002

In this Paper, we Propose model-based tuning rules of the PID controller incorporating with genetic algorithms. Three sets of optimal PID parameters for step set-point tracking are obtained based on the first-order time delay model of plants and a genetic algorithm which minimizes performance indices(IAE, ISE and ITAE). Then tuning rules are obtained using the tuned parameter sets, potential rule models and a genetic algorithm. Simulation is carried out to verify the effectiveness of the proposed rules.

• Journal of Institute of Control, Robotics and Systems
• /
• v.12 no.7
• /
• pp.644-649
• /
• 2006

In this paper, a methodology for estimating the model parameters of a discrete-time system and tuning a digital PID controller based on the estimated model and a genetic algorithm is presented. To deal with optimization problems regarding parameter estimation and controller tuning, pseudo-parallel genetic algorithms(PPGAs) are used. The parameters of a discrete-time system are estimated using both the model adjustment technique and a PPGA. The digital PID controller is described by the pulse transfer function and then its three gains are tuned based on both the model reference technique and another PPGA. A set of experimental works on two processes are carried out to illustrate the performance of the proposed method.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.19 no.6
• /
• pp.797-803
• /
• 2010

A method has been developed to tune the static stiffness at a rotation joint considering the whole machine tool system by interactive use of finite element method and experiment. This paper describes the procedure of this method and shows the results. The method uses the static experiment on measurement model which is set-up so that the effects of uncertain factors can be excluded. For FEM simulation, the rotation joint model is simplified using only spindle, bearing and spring. At the rotation joint, the damping coefficient is ignored, The spindle and bearing is connected by only spring. By static experiment, 500 N is forced to the front and behind portion of spindle and the deformation is measured by capacitive sensor. The deformation by FEM simulation is extracted with changing the static stiffness from the initial static stiffness considering only rotation joint. The tuning static stiffness is obtained by exploring the static stiffness directly trusting the deformation from the static experiment. Finally, the general tuning method of the static stiffness of machine tool joint is proposed using the force stream and the modal analysis of machine tool.

• Proceedings of the KIEE Conference
• /
• 2007.04a
• /
• pp.207-209
• /
• 2007

In this paper, we introduce a genetic optimization of fuzzy set-fuzzy model using successive tuning method to carry out the model identification of complex and nonlinear systems. To identity we use genetic alrogithrt1 (GA) sand C-Means clustering. GA is used for determination the number of input, the seleced input variables, the number of membership function, and the conclusion inference type. Information Granules (IG) with the aid of C-Means clustering algorithm help determine the initial paramters of fuzzy model such as the initial apexes of the, membership functions in the premise part and the initial values of polyminial functions in the consequence part of the fuzzy rules. The overall design arises as a hybrid structural and parametric optimization. Genetic algorithms and C-Means clustering are used to generate the structurally as well as parametrically optimized fuzzy model. To identify the structure and estimate parameters of the fuzzy model we introduce the successive tuning method with variant generation-based evolution by means of GA. Numerical example is included to evaluate the performance of the proposed model.