• 제목/요약/키워드: Weighted Loss Function

검색결과 51건 처리시간 0.024초

Sufficient Conditions for the Admissibility of Estimators in the Multiparameter Exponential Family

  • Dong, Kyung-Hwa;Kim, Byung-Hwee
    • Journal of the Korean Statistical Society
    • /
    • 제22권1호
    • /
    • pp.55-69
    • /
    • 1993
  • Consider the problem of estimating an arbitrary continuous vector function under a weighted quadratic loss in the multiparameter exponential family with the density of the natural form. We first provide, using Blyth's (1951) method, a set of sufficient conditions for the admisibility of (possibly generalized Bayes) estimators and then treat some examples for normal, Poisson, and gamma distributions as applications of the main result.

  • PDF

Improvement of learning concrete crack detection model by weighted loss function

  • Sohn, Jung-Mo;Kim, Do-Soo;Hwang, Hye-Bin
    • 한국컴퓨터정보학회논문지
    • /
    • 제25권10호
    • /
    • pp.15-22
    • /
    • 2020
  • 본 연구에서는 가중치 오차 함수를 적용하여, 미세한 콘크리트 균열을 감지하는 U-Net 모델을 만들 수 있도록 개선 방안을 제안한다. 콘크리트 균열은 안전을 위협하는 요소이기 때문에 그 상태를 주기적으로 파악하고 신속하게 초기 대응을 하는 것이 중요하다. 하지만 현재는 점검자가 직접 육안으로 검사하고 평가하는 외관 검사법이 주로 사용되고 있다. 이는 정확성뿐만 아니라 비용과 시간, 안전성 측면에서도 한계점을 가진다. 이에 콘크리트 구조물에 생성되는 미세한 균열을 신속하고 정밀하게 탐지할 수 있도록 딥러닝을 활용한 기술들이 연구되고 있다. 본 연구에서 U-Net을 활용한 균열 탐지를 시도한 결과, 미세한 균열을 탐지하지 못하는 것을 확인하였다. 이에 제시한 가중치 오차 함수를 적용하여 학습한 모델에 대해 성능을 검증한 결과, 정확도(Accuracy) 99% 이상, 조화평균(F1_Score) 89%에서 92%의 신뢰성 높은 수치를 도출해내었고, 미세한 균열을 정확하고 선명하게 탐지한 결과를 통해 학습 개선 방안의 성능을 검증하였다.

주파수 가중함수를 적용한 흡기계의 강건설계 연구 (Study on the Robust Design of an Intake System Using a Frequency Weighting Function)

  • 이종규;박영원;채장범
    • 한국소음진동공학회논문집
    • /
    • 제15권6호
    • /
    • pp.680-686
    • /
    • 2005
  • This paper introduces the robust design of an intake system using transmission loss and the frequency weighting function. First, transmission loss is measured to evaluate the performance of the noise reduction for the intake system. The robust design parameters of the intake system are extracted by adapting a cost function with the Taguchi method. Subsequently, the frequency weighting function is developed by the subjective evaluation in which 6 special engineers were participated. Finally, the comparison between the proposed frequency weighted optimal design and unweighted optimal design for the transmission loss as the part is performed. Here, the overall levels of the transmission loss according to the methods are presented to validate the effectiveness of the proposed methodology.

A Comparative Study for Several Bayesian Estimators Under Squared Error Loss Function

  • Kim, Yeong-Hwa
    • Journal of the Korean Data and Information Science Society
    • /
    • 제16권2호
    • /
    • pp.371-382
    • /
    • 2005
  • The paper compares the performance of some widely used Bayesian estimators such as Bayes estimator, empirical Bayes estimator, constrained Bayes estimator and constrained Bayes estimator by means of a new measurement under squared error loss function for the typical normal-normal situation. The proposed measurement is a weighted sum of the precisions of first and second moments. As a result, one can gets the criterion according to the size of prior variance against the population variance.

  • PDF

A Modification of the Combined Estimator of Inter- and Intra-Block Estimators under an Arbitrary Convex Loss Function

  • Lee, Young-Jo
    • Journal of the Korean Statistical Society
    • /
    • 제16권1호
    • /
    • pp.21-25
    • /
    • 1987
  • The combined estimator of inter- and intra-block estimators in incomplete block designs can be expressed as a weighted average of two location estimators. The weight should be between 0 and 1. However, the negative variance component estimate could result in the weight being negative or larger than 1. In this paper, we show that if two location estimators have symmetric unimodal distributions, truncating the weight to 0 or 1 accordingly improves the combined estimator under an arbitrary convex loss function.

  • PDF

가중치 손실 함수를 가지는 순환 컨볼루션 신경망 기반 주가 예측 (A Stock Price Prediction Based on Recurrent Convolution Neural Network with Weighted Loss Function)

  • 김현진;정연승
    • 정보처리학회논문지:소프트웨어 및 데이터공학
    • /
    • 제8권3호
    • /
    • pp.123-128
    • /
    • 2019
  • 본 논문에서는 RCNN (recurrent convolution neural network) 계층 모델을 채택한 인공 지능에 기반을 둔 주가 예측을 제안한다. LSTM (long-term memory model) 기반 신경망은 시계열 데이터의 예측에 사용된다. 다른 한편, 컨볼루션 신경망은 데이터 필터링, 평균화 및 데이터 확장을 제공한다. 제안된 주가 예측에서는 위에서 언급 한 장점들을 RCNN 모델에서 결합하여 적용함으로써 다음날의 주가 종가를 예측한다. 그리고 최근의 시계열의 데이터를 강조하기 위해 커스텀 가중치 손실 함수가 채택되었다. 또한 시장의 상황을 반영하기 위해 주가 인덱스에 관련된 데이터를 입력으로 포함하였다. 제안된 주가 예측 방식은 실제 주가를 대상으로 한 실험에서 3.19%로 테스트 오차를 줄였으며, 다른 방법보다 약 19%의 성능 향상을 거둘 수 있었다.

A Novel Algorithm of Joint Probability Data Association Based on Loss Function

  • Jiao, Hao;Liu, Yunxue;Yu, Hui;Li, Ke;Long, Feiyuan;Cui, Yingjie
    • KSII Transactions on Internet and Information Systems (TIIS)
    • /
    • 제15권7호
    • /
    • pp.2339-2355
    • /
    • 2021
  • In this paper, a joint probabilistic data association algorithm based on loss function (LJPDA) is proposed so that the computation load and accuracy of the multi-target tracking algorithm can be guaranteed simultaneously. Firstly, data association is divided in to three cases based on the relationship among validation gates and the number of measurements in the overlapping area for validation gates. Also the contribution coefficient is employed for evaluating the contribution of a measurement to a target, and the loss function, which reflects the cost of the new proposed data association algorithm, is defined. Moreover, the equation set of optimal contribution coefficient is given by minimizing the loss function, and the optimal contribution coefficient can be attained by using the Newton-Raphson method. In this way, the weighted value of each target can be achieved, and the data association among measurements and tracks can be realized. Finally, we compare performances of LJPDA proposed and joint probabilistic data association (JPDA) algorithm via numerical simulations, and much attention is paid on real-time performance and estimation error. Theoretical analysis and experimental results reveal that the LJPDA algorithm proposed exhibits small estimation error and low computation complexity.

Variable selection in censored kernel regression

  • Choi, Kook-Lyeol;Shim, Jooyong
    • Journal of the Korean Data and Information Science Society
    • /
    • 제24권1호
    • /
    • pp.201-209
    • /
    • 2013
  • For censored regression, it is often the case that some input variables are not important, while some input variables are more important than others. We propose a novel algorithm for selecting such important input variables for censored kernel regression, which is based on the penalized regression with the weighted quadratic loss function for the censored data, where the weight is computed from the empirical survival function of the censoring variable. We employ the weighted version of ANOVA decomposition kernels to choose optimal subset of important input variables. Experimental results are then presented which indicate the performance of the proposed variable selection method.

Application of YOLOv5 Neural Network Based on Improved Attention Mechanism in Recognition of Thangka Image Defects

  • Fan, Yao;Li, Yubo;Shi, Yingnan;Wang, Shuaishuai
    • KSII Transactions on Internet and Information Systems (TIIS)
    • /
    • 제16권1호
    • /
    • pp.245-265
    • /
    • 2022
  • In response to problems such as insufficient extraction information, low detection accuracy, and frequent misdetection in the field of Thangka image defects, this paper proposes a YOLOv5 prediction algorithm fused with the attention mechanism. Firstly, the Backbone network is used for feature extraction, and the attention mechanism is fused to represent different features, so that the network can fully extract the texture and semantic features of the defect area. The extracted features are then weighted and fused, so as to reduce the loss of information. Next, the weighted fused features are transferred to the Neck network, the semantic features and texture features of different layers are fused by FPN, and the defect target is located more accurately by PAN. In the detection network, the CIOU loss function is used to replace the GIOU loss function to locate the image defect area quickly and accurately, generate the bounding box, and predict the defect category. The results show that compared with the original network, YOLOv5-SE and YOLOv5-CBAM achieve an improvement of 8.95% and 12.87% in detection accuracy respectively. The improved networks can identify the location and category of defects more accurately, and greatly improve the accuracy of defect detection of Thangka images.

어린이 음성인식을 위한 동적 가중 손실 기반 도메인 적대적 훈련 (Dynamically weighted loss based domain adversarial training for children's speech recognition)

  • 마승희
    • 한국음향학회지
    • /
    • 제41권6호
    • /
    • pp.647-654
    • /
    • 2022
  • 어린이 음성인식의 활용 분야가 증가하고 있지만, 양질의 데이터 부족은 어린이 음성인식 성능 향상의 걸림돌이 되고 있다. 본 논문은 성인의 음성 데이터를 추가로 사용하여 어린이 음성인식 성능을 개선하는 방법을 새롭게 제안한다. 제안하는 방법은 성인 학습 데이터양이 증가할수록 커지는 연령 간 데이터 불균형을 효과적으로 다루기 위해 dynamically weighted loss를 사용하여 트랜스포머 기반 도메인 적대적 훈련하는 방식이다. 구체적으로, 학습 중 미니 배치 내 클래스 불균형 정도를 수치화하고, 데이터가 적을수록 큰 가중치를 갖도록 손실함수를 정의하여 사용하였다. 실험에서는 성인과 어린이 학습 데이터 간 비대칭성에 따른 제안된 도메인 적대적 훈련의 효용성을 검증하였다. 실험 결과, 학습 데이터 내 연령 간 비대칭이 발생하는 모든 조건에서 제안하는 방법이 기존 도메인 적대적 훈련 방식보다 높은 어린이 음성인식 성능을 가짐을 확인할 수 있었다.