• 한국국방경영분석학회지
• /
• 제36권2호
• /
• pp.1-10
• /
• 2010

Learning rate is generally applied to estimate an appropriate production labor cost. Learning effect is obtained from repetitive work during the production period under 3 assumptions ; homogeneous production, same producer, quantity measure in continuous unit. However, production breaks occur frequently in Korean defense industry environment because of budget constraint and annual requirements. In this case previous learning effect can not be applied due to learning loss. This paper proposed the application of learning rate when a production break occurs in Korea defense industry. To obtain a learning loss, we surveyed various learning loss factors for different production breaks(6, 12, 18 months) from 4 defense industry companies. Then, we estimate the first unit labor hours in re-production phase after production break using Anderlohr method and Retrograde method with the result of the survey. This work is the first attempt to show a method which defines and evaluates the learning loss factors in Korean defense industry environment.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 춘계학술대회
• /
• pp.72-74
• /
• 2021

다양한 분야에서 현재 활용되고 있는 딥러닝 과정은 데이터 준비, 데이터 전처리, 모델 생성, 모델 학습, 모델 평가로 구성 된다. 이중 모델 학습 과정에서 손실함수는 모델이 학습하면서 출력한 값을 실제 값과 비교하여 그 차이를 출력하게 되고, 출력된 손실값을 기반으로 모델은 역전파 알고리즘을 통해 손실값이 감소하는 방향으로 가중치를 수정해가며 학습을 진행한다. 본 논문에서는 바이오마커 추출을 위한 딥러닝 모델에서 사용될 신경망 출력 값의 손실도를 측정하여 출력해주는 다양한 손실함수를 분석하고 실험을 통해 최적의 손실함수를 찾아내고자 한다.

• 반도체디스플레이기술학회지
• /
• 제22권2호
• /
• pp.40-44
• /
• 2023

The defect detection based on deep learning requires minimal loss and high accuracy to pinpoint product defects. In this paper, we confirm the loss rate of deep learning training based on disc-shaped artificial scaffold images. It is intended to compare the performance of Cross-Entropy functions used in object detection algorithms. The model was constructed using normal, defective artificial scaffold images and category cross entropy and sparse category cross entropy. The data was repeatedly learned five times using each loss function. The average loss rate, average accuracy, final loss rate, and final accuracy according to the loss function were confirmed.

• 한국컴퓨터정보학회논문지
• /
• 제29권2호
• /
• pp.1-12
• /
• 2024

경로 손실(Path Loss)을 예측하는 것은 셀룰러 네트워크(Cellular Network)에서 기지국(Base Station) 의 설치 위치 선정 등 무선망 설계에 중요한 요인 중 하나다. 기존에는 기지국의 최적 설치 위치를 결정하기 위해 수많은 현장 테스트(Field Tests)를 통해 경로 손실 값을 측정했다. 따라서 측정에 많은 시간이 소요된다는 단점이 있었다. 이러한 문제를 해결하기 위해 본 연구에서는 머신러닝(Machine Learning, ML) 기반의 경로 손실 예측 방법을 제안한다. 특히, 경로 손실 예측 성능을 향상시키기 위해서 앙상블 학습(Ensemble Learning) 접근법을 적용하였다. 부트스트랩 데이터 세트(Bootstrap Dataset)을 활용하여 서로 다른 하이퍼파라미터(Hyperparameter) 구성을 갖는 모델들을 얻고, 이 모델들을 앙상블하여 최종 모델을 구축했다. 인터넷상에 공개된 경로 손실 데이터 세트를 활용하여 제안하는 앙상블 기반 경로 손실 예측 방법과 다양한 ML 기반 방법들의 성능을 평가 및 비교했다. 실험 결과, 제안하는 방법이 기존 방법들보다 우수한 성능을 달성하였으며, 경로 손실 값을 가장 정확하게 예측할 수 있다는 것을 입증하였다.

• 한국멀티미디어학회논문지
• /
• 제24권10호
• /
• pp.1380-1390
• /
• 2021

The video face recognition (FR) is one of the most popular researches in the field of computer vision due to a variety of applications. In particular, research using the attention mechanism is being actively conducted. In video face recognition, attention represents where to focus on by using the input value of the whole or a specific region, or which frame to focus on when there are many frames. In this paper, we propose a novel attention based deep learning method. Main novelties of our method are (1) the use of combining two loss functions, namely weighted Softmax loss function and a Triplet loss function and (2) the feasibility of end-to-end learning which includes the feature embedding network and attention weight computation. The feature embedding network has a positive effect on the attention weight computation by using combined loss function and end-to-end learning. To demonstrate the effectiveness of our proposed method, extensive and comparative experiments have been carried out to evaluate our method on IJB-A dataset with their standard evaluation protocols. Our proposed method represented better or comparable recognition rate compared to other state-of-the-art video FR methods.

• 한국정보통신학회논문지
• /
• 제25권8호
• /
• pp.1046-1052
• /
• 2021

4차 산업의 발전과 고성능의 컴퓨팅 환경 구축으로 다양한 산업분야에서 인공지능이 적용되고 있다. 의료분야에서는 X-Ray, MRI, PET 등의 의료 영상 및 임상 자료를 이용하여 암, COVID-19, 골 연령 측정 등의 딥 러닝 학습이 진행되었다. 또한 스마트 의료기기, IoT 디바이스와 딥 러닝 알고리즘을 적용하여 ICT 의료 융합 기술 등이 연구되고 있다. 이러한 기술 중 의료 영상 기반 딥 러닝 학습은 의료 영상의 바이오마커를 정확히 찾아내고, 최소한의 손실률과 높은 정확도가 필요하다. 따라서 본 논문은 흉부 X-Ray 이미지 기반 딥 러닝 학습 과정에서 손실률을 도출하는 손실 함수 중 영상분류 알고리즘에서 사용되는 Cross-Entropy 함수들의 성능을 비교·분석하고자 한다.

• International Journal of Computer Science & Network Security
• /
• 제23권6호
• /
• pp.35-48
• /
• 2023

The study aimed to provide a proposed conception of technical solutions to address the educational loss in mathematics in the fifth grade of primary school, and the study adopted the descriptive approach. male and female teachers, and to achieve the study objective, the researcher built a tool for the study, a "questionnaire", in which he used the comprehensive inventory method, and the results of the study showed: identifying the learning outcomes that represent an educational loss, and identifying the learning outcomes that are considered essential in teaching and learning mathematics for the fifth grade of primary school. In the event that it is not achieved by students, it is considered an educational loss that may affect the future of students' education and learning. Because it is a basis for later experiences in mathematics in other classes, and the study also found the effectiveness of the proposed visualization of technical solutions provided to address the educational loss in mathematics for the fifth grade: (short electronic tests, YouTube channel, homework, educational platform, electronic worksheets, and communication channels).

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권3호
• /
• pp.670-684
• /
• 2024

3D Cross-Modal Retrieval (3DCMR) is a task that retrieves 3D objects regardless of modalities, such as images, meshes, and point clouds. One of the most prominent methods used for 3DCMR is the Cross-Modal Center Loss Function (CLF) which applies the conventional center loss strategy for 3D cross-modal search and retrieval. Since CLF is based on center loss, the center features in CLF are also susceptible to subtle changes in hyperparameters and external inferences. For instance, performance degradation is observed when the batch size is too small. Furthermore, the Mean Squared Error (MSE) used in CLF is unable to adapt to changes in batch size and is vulnerable to data variations that occur during actual inference due to the use of simple Euclidean distance between multi-modal features. To address the problems that arise from small batch training, we propose a Noisy Center Loss (NCL) method to estimate the optimal center features. In addition, we apply the simple Siamese representation learning method (SimSiam) during optimal center feature estimation to compare projected features, making the proposed method robust to changes in batch size and variations in data. As a result, the proposed approach demonstrates improved performance in ModelNet40 dataset compared to the conventional methods.

• 대한원격탐사학회지
• /
• 제36권5_2호
• /
• pp.929-937
• /
• 2020

본 논문에서는 의미론적 분할을 위한 딥러닝 기술 중의 하나인 UNet++ 모델을 이용하여 다시기 위성영상의 변화지역을 탐지하고자 하였다. 다양한 손실함수에 대한 학습결과를 분석하기 위하여, 이진 교차 엔트로피, 자카드 변수에 의하여 학습된 UNet++ 모델에 의한 변화탐지 결과를 평가하였다. 또한, 딥러닝 모델의 결과는 WorldView-3 위성영상을 활용하여 기존의 화소기반 변화탐지 기법의 결과와 비교하여 평가하였다. 실험결과, 손실함수의 특성에 따라서 딥러닝 모델의 성능이 달라질 수 있음을 확인하였으나, 기존 기법들과 비교하여 우수한 결과를 나타내는 것도 확인하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권6호
• /
• pp.1743-1758
• /
• 2023

Automatic modulation classification is a critical algorithm for non-cooperative communication systems. This paper addresses the challenging problem of closed-set and open-set signal modulation classification in complex channels. We propose a novel approach that incorporates a self-learning filter and center-loss in Deep Residual Shrinking Networks (DRSN) for closed-set modulation classification, and the Opendistance method for open-set modulation classification. Our approach achieves better performance than existing methods in both closed-set and open-set recognition. In closed-set recognition, the self-learning filter and center-loss combination improves recognition performance, with a maximum accuracy of over 92.18%. In open-set recognition, the use of a self-learning filter and center-loss provide an effective feature vector for open-set recognition, and the Opendistance method outperforms SoftMax and OpenMax in F1 scores and mean average accuracy under high openness. Overall, our proposed approach demonstrates promising results for automatic modulation classification, providing better performance in non-cooperative communication systems.