• 제목/요약/키워드: Squared error loss

검색결과 71건 처리시간 0.022초

3D Cross-Modal Retrieval Using Noisy Center Loss and SimSiam for Small Batch Training

  • Yeon-Seung Choo;Boeun Kim;Hyun-Sik Kim;Yong-Suk Park
    • 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.

잡음 환경에 효과적인 마스크 기반 음성 향상을 위한 손실함수 조합에 관한 연구 (A study on combination of loss functions for effective mask-based speech enhancement in noisy environments)

  • 정재희;김우일
    • 한국음향학회지
    • /
    • 제40권3호
    • /
    • pp.234-240
    • /
    • 2021
  • 본 논문에서는 잡음 환경에서 효과적인 음성 인식을 위해 마스크 기반의 음성 향상 기법을 개선한다. 마스크 기반의 음성 향상 기법에서는 심층 신경망을 기반으로 추정한 마스크를 잡음 오염 음성에 곱하여 향상된 음성을 얻는다. 마스크 추정 모델로 VoiceFilter(VF) 모델을 사용하고 추정된 마스크로 얻은 음성으로부터 잔여 잡음을 보다 확실히 제거하기 위해 Spectrogram Inpainting(SI)기법을 적용한다. 본 논문에서는 음성 향상 결과를 보다 개선하기 위해 마스크 추정을 위한 모델 학습 과정에 사용되는 조합된 손실함수를 제안한다. 음성 구간에 남아 있는 잡음을 보다 효과적으로 제거하기 위해 잡음 오염 음성에 마스크를 적용한 Triplet 손실함수의 Positive 부분을 컴포넌트 손실함수와 조합하여 사용한다. 실험 평가를 위한 잡음 음성 데이터는 TIMIT 데이터베이스와 NOISEX92, 배경음악 잡음을 다양한 Signal to Noise Ratio(SNR) 조건으로 합성하여 만들어 사용한다. 음성 향상의 성능 평가는 Source to Distortion Ratio(SDR), Perceptual Evaluation of Speech Quality(PESQ), Short-Time Objective Intelligibility(STOI)를 이용한다. 실험을 통해 평균 제곱 오차로만 훈련된 기존 시스템과 비교하여, VF 모델은 평균 제곱 오차로 훈련하고 SI 모델은 조합된 손실함수를 사용하였을 때 SDR은 평균 0.5dB, PESQ는 평균 0.06, STOI는 평균 0.002만큼 성능이 향상된 것을 확인했다.

Bayesian and maximum likelihood estimations from exponentiated log-logistic distribution based on progressive type-II censoring under balanced loss functions

  • Chung, Younshik;Oh, Yeongju
    • Communications for Statistical Applications and Methods
    • /
    • 제28권5호
    • /
    • pp.425-445
    • /
    • 2021
  • A generalization of the log-logistic (LL) distribution called exponentiated log-logistic (ELL) distribution on lines of exponentiated Weibull distribution is considered. In this paper, based on progressive type-II censored samples, we have derived the maximum likelihood estimators and Bayes estimators for three parameters, the survival function and hazard function of the ELL distribution. Then, under the balanced squared error loss (BSEL) and the balanced linex loss (BLEL) functions, their corresponding Bayes estimators are obtained using Lindley's approximation (see Jung and Chung, 2018; Lindley, 1980), Tierney-Kadane approximation (see Tierney and Kadane, 1986) and Markov Chain Monte Carlo methods (see Hastings, 1970; Gelfand and Smith, 1990). Here, to check the convergence of MCMC chains, the Gelman and Rubin diagnostic (see Gelman and Rubin, 1992; Brooks and Gelman, 1997) was used. On the basis of their risks, the performances of their Bayes estimators are compared with maximum likelihood estimators in the simulation studies. In this paper, research supports the conclusion that ELL distribution is an efficient distribution to modeling data in the analysis of survival data. On top of that, Bayes estimators under various loss functions are useful for many estimation problems.

A new extension of Lindley distribution: modified validation test, characterizations and different methods of estimation

  • Ibrahim, Mohamed;Yadav, Abhimanyu Singh;Yousof, Haitham M.;Goual, Hafida;Hamedani, G.G.
    • Communications for Statistical Applications and Methods
    • /
    • 제26권5호
    • /
    • pp.473-495
    • /
    • 2019
  • In this paper, a new extension of Lindley distribution has been introduced. Certain characterizations based on truncated moments, hazard and reverse hazard function, conditional expectation of the proposed distribution are presented. Besides, these characterizations, other statistical/mathematical properties of the proposed model are also discussed. The estimation of the parameters is performed through different classical methods of estimation. Bayes estimation is computed under gamma informative prior under the squared error loss function. The performances of all estimation methods are studied via Monte Carlo simulations in mean square error sense. The potential of the proposed model is analyzed through two data sets. A modified goodness-of-fit test using the Nikulin-Rao-Robson statistic test is investigated via two examples and is observed that the new extension might be used as an alternative lifetime model.

SOME POINT ESTIMATES FOR THE SHAPE PARAMETERS OF EXPONENTIATED-WEIBULL FAMILY

  • Singh Umesh;Gupta Pramod K.;Upadhyay S.K.
    • Journal of the Korean Statistical Society
    • /
    • 제35권1호
    • /
    • pp.63-77
    • /
    • 2006
  • Maximum product of spacings estimator is proposed in this paper as a competent alternative of maximum likelihood estimator for the parameters of exponentiated-Weibull distribution, which does work even when the maximum likelihood estimator does not exist. In addition, a Bayes type estimator known as generalized maximum likelihood estimator is also obtained for both of the shape parameters of the aforesaid distribution. Though, the closed form solutions for these proposed estimators do not exist yet these can be obtained by simple appropriate numerical techniques. The relative performances of estimators are compared on the basis of their relative risk efficiencies obtained under symmetric and asymmetric losses. An example based on simulated data is considered for illustration.

Bayesian estimation for the exponential distribution based on generalized multiply Type-II hybrid censoring

  • Jeon, Young Eun;Kang, Suk-Bok
    • Communications for Statistical Applications and Methods
    • /
    • 제27권4호
    • /
    • pp.413-430
    • /
    • 2020
  • The multiply Type-II hybrid censoring scheme is disadvantaged by an experiment time that is too long. To overcome this limitation, we propose a generalized multiply Type-II hybrid censoring scheme. Some estimators of the scale parameter of the exponential distribution are derived under a generalized multiply Type-II hybrid censoring scheme. First, the maximum likelihood estimator of the scale parameter of the exponential distribution is obtained under the proposed censoring scheme. Second, we obtain the Bayes estimators under different loss functions with a noninformative prior and an informative prior. We approximate the Bayes estimators by Lindleys approximation and the Tierney-Kadane method since the posterior distributions obtained by the two priors are complicated. In addition, the Bayes estimators are obtained by using the Markov Chain Monte Carlo samples. Finally, all proposed estimators are compared in the sense of the mean squared error through the Monte Carlo simulation and applied to real data.

미래손실에 기초한 통합공정관리계획 (An Integrated Process Control Scheme Based on the Future Loss)

  • 박창순;이재헌
    • 응용통계연구
    • /
    • 제21권2호
    • /
    • pp.247-264
    • /
    • 2008
  • 통합공정관리의 기본절차는 잡음이 내재하는 공정에 대하여 수정조치를 취하고, 수정활동 중 공정에 이상원인이 발생하면 관리도를 통하여 발생을 탐지하고 교정활동을 통하여 이를 제거하게 된다. 그러나 공정의 교정활동은 많은 시간과 비용을 수반하는 비생산적 요인을 유발할 수 있기 때문에 무조건적 교정활동은 생산성을 저하시키는 반대 급부도 동시에 내포하고 있다. 이 논문에서는 공정모형으로 ARIMA(0,1,1) 모형을 가정하고 공정 평균과 분산에 이상원인이 발생하는 경우 이를 탐지하는 절차를 소개하고, 이상신호의 시점에서 공정 잔여시간 동안 발생할 수 있는 미래손실에 기초하여 교정 활동의 여부를 판단하는 통합공정관리 절차를 제안한다.

신경 망의 지도 학습을 위한 로그 간격의 학습 자료 구성 방식과 손실 함수의 성능 평가 (Performance Evaluation of Loss Functions and Composition Methods of Log-scale Train Data for Supervised Learning of Neural Network)

  • 송동규;고세헌;이효민
    • Korean Chemical Engineering Research
    • /
    • 제61권3호
    • /
    • pp.388-393
    • /
    • 2023
  • 지도 학습 기반의 신경 망을 활용한 공학적 자료의 분석은 화학공학 공정 최적화, 미세 먼지 농도 추정, 열역학적 상평형 예측, 이동 현상 계의 물성 예측 등 다양한 분야에서 활용되고 있다. 신경 망의 지도 학습은 학습 자료를 요구하며, 주어진 학습 자료의 구성에 따라 학습 성능이 영향을 받는다. 빈번히 관찰되는 공학적 자료 중에는 DNA의 길이, 분석 물질의 농도 등과 같이 로그 간격으로 주어지는 자료들이 존재한다. 본 연구에서는 넓은 범위에 분포된 로그 간격의 학습 자료를 기계 학습으로 처리하는 경우, 사용 가능한 손실 함수들의 학습 성능을 정량적으로 평가하였으며, 적합한 학습 자료 구성 방식을 연구하였다. 이를 수행하고자, 100×100의 가상 이미지를 활용하여 기계 학습의 회귀 과업을 구성하였다. 4개의 손실 함수들에 대하여 (i) 오차 행렬, (ii) 최대 상대 오차, (iii) 평균 상대 오차로 정량적 평가하여, mape 혹은 msle가 본 연구에서 다룬 과업에 대해 최적의 손실 함수가 됨을 알아내었다. 또한, 학습 자료의 값이 넓은 범위에 걸쳐 분포하는 경우, 학습 자료의 구성을 로그 간격 등을 고려하여 균등 선별하는 방식이 높은 학습 성능을 보임을 밝혀내었다. 본 연구에서 다룬 회귀 과업은 DNA의 길이 예측, 생체 유래 분자 분석, 콜로이드 용액의 농도 추정 등의 공학적 과업에 적용 가능하며, 본 결과를 활용하여 기계 학습의 성능과 학습 효율의 증대를 기대할 수 있을 것이다.

RELIABILITY ANALYSIS FOR THE TWO-PARAMETER PARETO DISTRIBUTION UNDER RECORD VALUES

  • Wang, Liang;Shi, Yimin;Chang, Ping
    • Journal of applied mathematics & informatics
    • /
    • 제29권5_6호
    • /
    • pp.1435-1451
    • /
    • 2011
  • In this paper the estimation of the parameters as well as survival and hazard functions are presented for the two-parameter Pareto distribution by using Bayesian and non-Bayesian approaches under upper record values. Maximum likelihood estimation (MLE) and interval estimation are derived for the parameters. Bayes estimators of reliability performances are obtained under symmetric (Squared error) and asymmetric (Linex and general entropy (GE)) losses, when two parameters have discrete and continuous priors, respectively. Finally, two numerical examples with real data set and simulated data, are presented to illustrate the proposed method. An algorithm is introduced to generate records data, then a simulation study is performed and different estimates results are compared.

Modeling pediatric tumor risks in Florida with conditional autoregressive structures and identifying hot-spots

  • Kim, Bit;Lim, Chae Young
    • Journal of the Korean Data and Information Science Society
    • /
    • 제27권5호
    • /
    • pp.1225-1239
    • /
    • 2016
  • We investigate pediatric tumor incidence data collected by the Florida Association for Pediatric Tumor program using various models commonly used in disease mapping analysis. Particularly, we consider Poisson normal models with various conditional autoregressive structure for spatial dependence, a zero-in ated component to capture excess zero counts and a spatio-temporal model to capture spatial and temporal dependence, together. We found that intrinsic conditional autoregressive model provides the smallest Deviance Information Criterion (DIC) among the models when only spatial dependence is considered. On the other hand, adding an autoregressive structure over time decreases DIC over the model without time dependence component. We adopt weighted ranks squared error loss to identify high risk regions which provides similar results with other researchers who have worked on the same data set (e.g. Zhang et al., 2014; Wang and Rodriguez, 2014). Our results, thus, provide additional statistical support on those identied high risk regions discovered by the other researchers.