• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.4
• /
• pp.177-186
• /
• 2019

In wireless sensor networks, the transmission success ratio would be decreased when the scale of the WSNs increased. To defeat this problem, we propose a multipath routing based on opportunistic routing for improving end-to-end reliability in large-scale wireless sensor networks. The proposed scheme exploits the advantages of existing opportunistic routing and achieves high end-to-end success ratio by branching like a multipath routing through local decision without information of the whole network. As a result of the simulation result, the proposed scheme shows a similar or higher end-to-end transmission success ratio and less energy consumption rather than the existing scheme.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2003.06a
• /
• pp.1789-1794
• /
• 2003

As a new approach to analyze shear behaviors in the shear plane and chip-tool friction behaviors in the chip-tool contact region during an end milling process, this paper introduces a method to transform an end milling process to an equivalent oblique cutting process. In this approach, varying undeformed chip thicknesses and cutting forces in the up-and down-end milling process are replaced with the equivalent ones of oblique cutting. Accordingly, in the current paper, the shear and friction characteristics of end milling operations, up- and down-end milling, have been analyzed based on the equivalent oblique cutting models. Two series of cutting tests, up- and down-end milling tests and the equivalent oblique cutting tests to that, have been carried out to verify the validity of the analyses. And using the results of cutting tests the cutting characteristics of the up- and down-end milling processes have been thoroughly investigated.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.600-601
• /
• 2024

본 논문은 RAG(Retrieval-Augmented Generation) End2End의 리소스(Resource) 과부하 문제를 해결하는 동시에 모델 성능을 향상 시키기 위해 PEFT(Parameters-Efficient Fine-Tuning)기술인 LoRA(Low Rank Adaptation)적용에 관한 연구이다. 본 논문에서는 RAG End2End 모델의 파라미터 값과 개수를 유지하면서, LRM(Low Rank Matrices)을 이용하여 추가적인 파라미터만을 미세 조정하는 방식으로, 전반적인 모델의 효율성을 극대화하는 방안을 제시하였다. 본 논문에서 다양한 도메인에 데이터 셋에 대한 제안 방식의 성능을 검증하고자 Conversation, Covid-19, News 데이터 셋을 사용하였다. 실험결과, 훈련에 필요한 파라미터의 크기가 약 6.4억개에서 180만개로 감소하였고, EM(Exact Match)점수가 유사하거나 향상되었다. 이는 LoRA를 통한 접근 법이 RAG End2End 모델의 효율성을 개선할 수 있는 효과적인 전략임을 증명하였다.

• Journal of the Korean Wood Science and Technology
• /
• v.18 no.2
• /
• pp.79-85
• /
• 1990

Traditional complex joints have used to a wide variety of wooden furniture construction. Dowel joint is the most popular joint s. However design of this joint to meet specified service condition has been hampered by a lack of proven design formulas which can be use to predict their strength. The object of this study is to investigate the withdrawal strength and effect of dowel diameters in wood and wood based materials. The obtained results were as follows; 1. The relationship between withdrawal strength and dowel diameter is found to be linear. 2. Withdrawal strength of medium density fiberboard and Sepetir in end-to-side joints is superior to Antiaris, particleboard and plywood. 3. In end-to-end joints, withdrawal strength of medium density fiberboard is the most superior joint. but Sepetir. Antiaris and plywood have similarly strength and particleboard is inferiority. 4. Withdrawal strength in end-to-end joints of Antiaris and plywood is higher than in end-ta-side joints. But in end-to-end joints of Sepetir. medium density fiberboard and particleboard is similarity in end-to-side joints.

• Journal of Communications and Networks
• /
• v.18 no.4
• /
• pp.580-588
• /
• 2016

Energy efficiency is the main objective in the design of a wireless sensor network (WSN). In many applications, sensing data must be transmitted from sources to a sink in a timely manner. This paper describes an investigation of the trade-off between two objectives in WSN design: minimizing energy consumption and minimizing end-to-end delay. We first propose a new distributed clustering approach to determining the best clusterhead for each cluster by considering both energy consumption and end-to-end delay requirements. Next, we propose a new energy-cost function and a new end-to-end delay function for use in an inter-cluster routing algorithm. We present a multi-hop routing algorithm for use in disseminating sensing data from clusterheads to a sink at the minimum energy cost subject to an end-to-end delay constraint. The results of a simulation are consistent with our theoretical analysis results and show that our proposed performs much better than similar protocols in terms of energy consumption and end-to-end delay.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2010.10a
• /
• pp.465-467
• /
• 2010

This paper proposes MIMO two-way cooperative relay scheme to optimize the end to end capacity in wireless multi-hop mesh network. The basic idea is to perform data transmission via multi-hop relay nodes, in equidistant topology, this method is quite efficient. However, on one hand this topology is very rare in practical situation, on the other hand, in real practical situation where the topology is most likely non equidistant, the end to end capacity significantly degrades due to bottleneck link caused by uneven SNR. Moreover, the end to end capacity degrades at high SNR due to overreach interference from far nodes existing in multi-hop relay networks. In this paper, MIMO two-way cooperative relay in the region of non equidistant nodes is found efficient to improve the end to end capacity. The proposed scheme is validated using numerical simulation.

• Geomechanics and Engineering
• /
• v.23 no.6
• /
• pp.523-533
• /
• 2020

The current study focuses on the effect of the end shape of steel pipe piles on installation effort and bearing resistance using the pressing method of installation under dense ground conditions. The effect of pile rotation on the installation effort and bearing resistance is also investigated. The model steel piles with a flat end, cone end and cutting-edge end were used in this study. The test results indicated that cone end pile with the pressing method of installation required the least installation effort (load) and showed higher ultimate resistance than flat and cutting-edge end piles. However, pressing and rotation during cutting-edge end pile installation considerably reduces the installation effort (load and torque) if pile penetration in one rotation equal to the cutting-edge depth. Inclusion of rotation during pile installation reduces the ultimate bearing resistance. However, if penetration of the cutting-edge end pile equal to the cutting-edge depth in one rotation, the reduction in ultimate resistance can be minimized. In comparing the cone and cutting-edge end piles installed with pressing and rotation, the least installation effort is observed in the cutting-edge end pile installed with penetration rate equal to the cutting-edge depth per rotation.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.259-263
• /
• 2021

상호참조해결은 주어진 문서에서 상호참조해결 대상이 되는 멘션(mention)을 식별하고, 동일한 개체(entity)를 의미하는 멘션들을 찾아 그룹화하는 자연어처리 태스크이다. 최근 상호참조해결에서는 BERT를 이용하여 단어의 문맥 표현을 얻은 후, 멘션 탐지와 상호참조해결을 동시에 진행하는 end-to-end 모델이 주로 연구되었으나, 512 토큰 이상의 긴 문서를 처리하기 위해서는 512 토큰 이하로 문서를 분할하여 처리하기 때문에 길이가 긴 문서에 대해서는 상호참조해결 성능이 낮아지는 문제가 있다. 본 논문에서는 512 토큰 이상의 긴 문서를 위한 BERT 기반의 end-to-end 상호참조해결 모델을 제안한다. 본 모델은 긴 문서를 512 이하의 토큰으로 쪼개어 기존의 BERT에서 단어의 1차 문맥 표현을 얻은 후, 이들을 다시 연결하여 긴 문서의 Global Positional Encoding 또는 Embedding 값을 더한 후 Global BERT layer를 거쳐 단어의 최종 문맥 표현을 얻은 후, end-to-end 상호참조해결 모델을 적용한다. 실험 결과, 본 논문에서 제안한 모델이 기존 모델과 유사한 성능을 보이면서(테스트 셋에서 0.16% 성능 향상), GPU 메모리 사용량은 1.4배 감소하고 속도는 2.1배 향상되었다.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• Phonetics and Speech Sciences
• /
• v.11 no.1
• /
• pp.41-49
• /
• 2019

An end-to-end speech recognition model consisting of a single integrated neural network model was recently proposed. The end-to-end model does not need several training steps, and its structure is easy to understand. However, it is difficult to understand how the model recognizes speech internally. In this paper, we visualized and analyzed the attention-based end-to-end model to elucidate its internal mechanisms. We compared the acoustic model of the BLSTM-HMM hybrid model with the encoder of the end-to-end model, and visualized them using t-SNE to examine the difference between neural network layers. As a result, we were able to delineate the difference between the acoustic model and the end-to-end model encoder. Additionally, we analyzed the decoder of the end-to-end model from a language model perspective. Finally, we found that improving end-to-end model decoder is necessary to yield higher performance.