• 제목/요약/키워드: Standard error of prediction

검색결과 324건 처리시간 0.026초

보정신경망을 이용한 냉연 압하력 적중율 향상 (Improvement of roll force precalculation accuracy in cold mill using a corrective neural network)

  • 이종영;조형석;조성준;조용중;윤성철
    • 제어로봇시스템학회:학술대회논문집
    • /
    • 제어로봇시스템학회 1996년도 한국자동제어학술회의논문집(국내학술편); 포항공과대학교, 포항; 24-26 Oct. 1996
    • /
    • pp.1083-1086
    • /
    • 1996
  • Cold rolling mill process in steel works uses stands of rolls to flatten a strip to a desired thickness. At cold rolling mill process, precalculation determines the mill settings before a strip actually enters the mill and is done by an outdated mathematical model. A corrective neural network model is proposed to improve the accuracy of the roll force prediction. Additional variables to be fed to the network include the chemical composition of the coil, its coiling temperature and the aggregated amount of processed strips of each roll. The network was trained using a standard backpropagation with 4,944 process data collected from no.1 cold rolling mill process from March 1995 through December 1995, then was tested on the unseen 1,586 data from Jan 1996 through April 1996. The combined model reduced the prediction error by 32.8% on average.

  • PDF

Discriminative Training of Sequence Taggers via Local Feature Matching

  • Kim, Minyoung
    • International Journal of Fuzzy Logic and Intelligent Systems
    • /
    • 제14권3호
    • /
    • pp.209-215
    • /
    • 2014
  • Sequence tagging is the task of predicting frame-wise labels for a given input sequence and has important applications to diverse domains. Conventional methods such as maximum likelihood (ML) learning matches global features in empirical and model distributions, rather than local features, which directly translates into frame-wise prediction errors. Recent probabilistic sequence models such as conditional random fields (CRFs) have achieved great success in a variety of situations. In this paper, we introduce a novel discriminative CRF learning algorithm to minimize local feature mismatches. Unlike overall data fitting originating from global feature matching in ML learning, our approach reduces the total error over all frames in a sequence. We also provide an efficient gradient-based learning method via gradient forward-backward recursion, which requires the same computational complexity as ML learning. For several real-world sequence tagging problems, we empirically demonstrate that the proposed learning algorithm achieves significantly more accurate prediction performance than standard estimators.

A Study on the Interrelationship between the Prediction Error and the Rating's Pattern in Collaborative Filtering

  • Lee, Seok-Jun;Kim, Sun-Ok;Lee, Hee-Choon
    • Journal of the Korean Data and Information Science Society
    • /
    • 제18권3호
    • /
    • pp.659-668
    • /
    • 2007
  • Collaborative filtering approach for recommender systems are now widely applied in e-commerce to assist customers to find their needs from many that are frequently available. this approach makes recommendations for users based on the opinions to similar users in the system. But this approach is opened to users who present their preference to items or acquire the preference information form other users, noise in the system makes significant problem for accurate recommendation. In this paper, we analyze the relationship between the standard deviation of preference ratings for each user and the estimated ratings of them. The result shows that the possibility of the pre-filtering condition which detecting the factor of bad effect on the prediction of user's preference. It is expected that using this result will reduce the possibility of bad effect on recommender systems.

  • PDF

지상용 초분광 스캐너를 활용한 사과의 당도예측 모델의 성능향상을 위한 연구 (Study of Prediction Model Improvement for Apple Soluble Solids Content Using a Ground-based Hyperspectral Scanner)

  • 송아람;전우현;김용일
    • 대한원격탐사학회지
    • /
    • 제33권5_1호
    • /
    • pp.559-570
    • /
    • 2017
  • 본 연구에서는 야외에서 자료 취득이 가능하며 한 번에 다량의 사과를 촬영할 수 있는 지상용 초분광 스캐너를 활용하여 사과의 분광정보와 당도와의 부분최소제곱회귀분석(PLSR, Partial Least Square Regression)을 수행하였으며, 최적의 예측모델을 구축하기 위한 다양한 전처리기법의 적용가능성을 평가하고 VIP(Variable Importance in Projection)점수를 통한 최적밴드를 산출하였다. 이를 위하여 360-1019 nm영역에서 촬영된 515밴드의 초분광 영상에서 70개의 분광곡선을 취득하였으며, 디지털광도계를 이용하여 당도($^{\circ}Brix$)를 측정하였다. 사과의 분광특성과 당도사이의 회귀모델을 구축하였으며, 최적의 예측모델은 모델 예측치와 실측치간의 결정계수($r_p^2$, coefficient of determination of prediction)와 RMSECV(Root Mean Square Error of Cross Validation), RMSEP(Root Mean Square Error of Prediction)등을 고려하여 선정하였다. 그 결과 산란보정 기법의 대표적인 MSC(Multiplicative Scatter Correction)의 기반의 전처리기법이 가장 효과적이었으며, MSC와 SNV(Standard Normal Variate)를 조합한 경우 RMSECV와 RMSEP가 각각 0.8551과 0.8561로 가장 낮았고, $r_c^2$$r_p^2$은 각각 0.8533과 0.6546으로 가장 높았다, 또한 360-380, 546-690, 760, 915, 931-939, 942, 953, 971, 978, 981, 988, 992-1019 nm 등이 당도 측정을 위한 가장 영향력 있는 파장영역으로 나타났다. 해당 영역의 분광값을 가지고 PLSR을 수행한 결과, 전파장대를 사용할 때보다 RMSEP가 0.6841로 감소하고 $r_p^2$는 0.7795로 증가하는 것을 확인하였다. 본 연구를 통하여 사과의 당도측정에 있어 야외에서 취득한 초분광 영상자료의 활용 가능성을 확인하였으며, 이는 필드자료 및 센서 활용분야의 확장가능성을 보여준다.

마이크로파 자유공간 전송을 이용한 산물벼 함수율 측정장치 개발 (Development of Moisture Content Measurement Device for Paddy Rice using Microwave Free Space Transmission)

  • 김기복;김종헌;노상하
    • Journal of Biosystems Engineering
    • /
    • 제24권3호
    • /
    • pp.235-242
    • /
    • 1999
  • This study was conducted to develop a grain moisture meter using microwave free space transmission technique at X-band frequency. The 10.5GHz microwave oscillator using a dielectric resonator was designed and fabricated to transmit electromagnetic wave through standard horn antenna to a sample holder with the wetted Hwasung and Chuchung rough rice(12.00∼26.25%). To detect the output voltage of transmitted wave from receiving horn antenna, the detector was composed of shottkey diode and RF impedance matching circuit. The regression model for measurement of grain moisture content was developed. Its correlation coefficient and standard error of prediction (SEP) were found to be 0.9882 and 0.657 respectively between measure and predicted moisture contents.

  • PDF

Research on prediction and analysis of supercritical water heat transfer coefficient based on support vector machine

  • Ma Dongliang;Li Yi;Zhou Tao;Huang Yanping
    • Nuclear Engineering and Technology
    • /
    • 제55권11호
    • /
    • pp.4102-4111
    • /
    • 2023
  • In order to better perform thermal hydraulic calculation and analysis of supercritical water reactor, based on the experimental data of supercritical water, the model training and predictive analysis of the heat transfer coefficient of supercritical water were carried out by using the support vector machine (SVM) algorithm. The changes in the prediction accuracy of the supercritical water heat transfer coefficient are analyzed by the changes of the regularization penalty parameter C, the slack variable epsilon and the Gaussian kernel function parameter gamma. The predicted value of the SVM model obtained after parameter optimization and the actual experimental test data are analyzed for data verification. The research results show that: the normalization of the data has a great influence on the prediction results. The slack variable has a relatively small influence on the accuracy change range of the predicted heat transfer coefficient. The change of gamma has the greatest impact on the accuracy of the heat transfer coefficient. Compared with the calculation results of traditional empirical formula methods, the trained algorithm model using SVM has smaller average error and standard deviations. Using the SVM trained algorithm model, the heat transfer coefficient of supercritical water can be effectively predicted and analyzed.

지하수위 예측을 위한 경사하강법과 화음탐색법의 결합을 이용한 다층퍼셉트론 성능향상 (Improvement of multi layer perceptron performance using combination of gradient descent and harmony search for prediction of ground water level)

  • 이원진;이의훈
    • 한국수자원학회논문집
    • /
    • 제55권11호
    • /
    • pp.903-911
    • /
    • 2022
  • 물을 공급하기 위한 자원 중 하나인 지하수는 다양한 자연적 요인에 의해 수위의 변동이 발생한다. 최근, 인공신경망을 이용하여 지하수위의 변동을 예측하는 연구가 진행되었다. 기존에는 인공신경망 연산자 중 학습에 영향을 미치는 Optimizer로 경사하강법(Gradient Descent, GD) 기반 Optimizer를 사용하였다. GD 기반 Optimizer는 초기 상관관계 의존성과 해의 비교 및 저장 구조 부재의 단점이 존재한다. 본 연구는 GD 기반 Optimizer의 단점을 개선하기 위해 GD와 화음탐색법(Harmony Search, HS)를 결합한 새로운 Optimizer인 Gradient Descent combined with Harmony Search(GDHS)를 개발하였다. GDHS의 성능을 평가하기 위해 다층퍼셉트론(Multi Layer Perceptron, MLP)을 이용하여 이천율현 관측소의 지하수위를 학습 및 예측하였다. GD 및 GDHS를 사용한 MLP의 성능을 비교하기 위해 Mean Squared Error(MSE) 및 Mean Absolute Error(MAE)를 사용하였다. 학습결과를 비교하면, GDHS는 GD보다 MSE의 최대값, 최소값, 평균값 및 표준편차가 작았다. 예측결과를 비교하면, GDHS는 GD보다 모든 평가지표에서 오차가 작은 것으로 평가되었다.

Extrapolation of wind pressure for low-rise buildings at different scales using few-shot learning

  • Yanmo Weng;Stephanie G. Paal
    • Wind and Structures
    • /
    • 제36권6호
    • /
    • pp.367-377
    • /
    • 2023
  • This study proposes a few-shot learning model for extrapolating the wind pressure of scaled experiments to full-scale measurements. The proposed ML model can use scaled experimental data and a few full-scale tests to accurately predict the remaining full-scale data points (for new specimens). This model focuses on extrapolating the prediction to different scales while existing approaches are not capable of accurately extrapolating from scaled data to full-scale data in the wind engineering domain. Also, the scaling issue observed in wind tunnel tests can be partially resolved via the proposed approach. The proposed model obtained a low mean-squared error and a high coefficient of determination for the mean and standard deviation wind pressure coefficients of the full-scale dataset. A parametric study is carried out to investigate the influence of the number of selected shots. This technique is the first of its kind as it is the first time an ML model has been used in the wind engineering field to deal with extrapolation in wind performance prediction. With the advantages of the few-shot learning model, physical wind tunnel experiments can be reduced to a great extent. The few-shot learning model yields a robust, efficient, and accurate alternative to extrapolating the prediction performance of structures from various model scales to full-scale.

H.264 비디오 표준에서의 칼만 필터 기반의 움직임벡터 복원 (Kalman filter based Motion Vector Recovery for H.264)

  • 고기홍;김성환
    • 정보처리학회논문지D
    • /
    • 제14D권7호
    • /
    • pp.801-808
    • /
    • 2007
  • MPEG-2, MPEG-4, H.263, H.264 와 같은 부호화 표준은 비디오 영상을 압축하여 대역폭이 제한된 유/무선 통신 시스템을 통하여 전송한다. 통신 시스템에서 고압축률의 비트스트림은 채널 잡음 (channel noise)에 민감하여, 채널 잡음으로 인한 오류가 발생하기 쉽다. 이러한 오류는 수신부에서 디코딩할 때 비디오 영상을 심각하게 왜곡시키게 된다. 본 논문에서는 수신부 단에서 오류를 복원하는 기법 (decoder error concealment) 중 손상된 움직임벡터를 복원하는 기법을 제안한다. 본 논문에서는 손실된 움직임벡터를 예측하기 위하여 인접 블록들의 움직임 벡터를, 예측필터의 일종인 칼만 필터의 입력 치로 사용하여, 손실된 움직임벡터의 최적 예상치를 만들어 손상된 움직임벡터를 복구하게 된다. H.264 비디오 코딩을 적용한 표준 테스트 영상에 대하여, 손실된 MVD (motion vector difference) 값을 0 으로 대체한 뒤, H.264 비디오 코딩에서 사용하고 있는 기본 움직임벡터 예측만을 사용한 경우와 본 논문에서 제안한 칼만 필터를 사용한 복원기법을 비교하였으며, 복원된 움직 임벡터와 원래 움직임벡터 값과의 차이를 나타내는 오차율을 비교한 결과 제안된 기법의 오차율이 평균 0.91 - 1.12 정도의 정확도가 향상된 것을 확인할 수 있다.

Use of Monte Carlo code MCS for multigroup cross section generation for fast reactor analysis

  • Nguyen, Tung Dong Cao;Lee, Hyunsuk;Lee, Deokjung
    • Nuclear Engineering and Technology
    • /
    • 제53권9호
    • /
    • pp.2788-2802
    • /
    • 2021
  • Multigroup cross section (MG XS) generation by the UNIST in-house Monte Carlo (MC) code MCS for fast reactor analysis using nodal diffusion codes is reported. The feasibility of the approach is quantified for two sodium fast reactors (SFRs) specified in the OECD/NEA SFR benchmark: a 1000 MWth metal-fueled SFR (MET-1000) and a 3600 MWth oxide-fueled SFR (MOX-3600). The accuracy of a few-group XSs generated by MCS is verified using another MC code, Serpent 2. The neutronic steady-state whole-core problem is analyzed using MCS/RAST-K with a 24-group XS set. Various core parameters of interest (core keff, power profiles, and reactivity feedback coefficients) are obtained using both MCS/RAST-K and MCS. A code-to-code comparison indicates excellent agreement between the nodal diffusion solution and stochastic solution; the error in the core keff is less than 110 pcm, the root-mean-square error of the power profiles is within 1.0%, and the error of the reactivity feedback coefficients is within three standard deviations. Furthermore, using the super-homogenization-corrected XSs improves the prediction accuracy of the control rod worth and power profiles with all rods in. Therefore, the results demonstrate that employing the MCS MG XSs for the nodal diffusion code is feasible for high-fidelity analyses of fast reactors.