基于DCNN和数据增强的固体发动机烧蚀预示方法

针对固体发动机烧蚀率的预示,现有传统建模方法存在复杂度高、计算需求大、试验数据少、样本不平衡等问题,提出了一种基于深度卷积神经网络和数据增强的固体发动机烧蚀率预示方法。将传感器数据处理为长度相同、特征相近的序列数据,并使用自适应高斯噪声和随机漂移这2种数据增强方法扩充数据样本,再将扩充后的试验样本和伪样本作为深度卷积神经网络的输入进行训练,将训练得到的模型与传统方法计算得到的烧蚀率预示值进行对比。结果表明,该方法下烧蚀率预示值误差低至0.013 5 m/s,预示精度可达95%。     

Exploring pixel-value differencing and base decomposition for low distortion data embedding

This paper presents a low distortion data embedding method using pixel-value differencing and base decomposition schemes. The pixel-value differencing scheme offers the advantage of conveying a large amount of payload, while still maintaining the consistency of an image characteristic after data embedding. We introduce the base decomposition scheme, which defines a base pair for each degree in order to construct a two-base notational system. This scheme provides the advantage of significantly reducing pixel variation encountered due to secret data embedding. We analyze the pixel variation and the expected mean square error caused by concealing with secret messages. The mathematical analysis shows that our scheme produces much smaller maximal pixel variations and expected mean square error while producing a higher PSNR. We evaluate the performance of our method using 6 categories of metrics which allow us to compare with seven other state-of-the-art algorithms. Experimental statistics verify that our algorithm outperforms existing counterparts in terms of lower image distortion and higher image quality. Finally, our scheme can survive from the RS steganalysis attack and the steganalytic histogram attack of pixel-value difference. We conclude that our proposed method is capable of embedding large amounts of a message, yet still produces the embedded image with very low distortion. To the best of our knowledge, in comparison with the current seven state-of-the-art data embedding algorithms, our scheme produces the lowest image distortion while embedding the same or slightly larger quantities of messages.     

New methods for imputation of missing genotype using linkage disequilibrium and haplotype information

In this paper, we propose new missing imputation methods for the missing genotype data of single nucleotide polymorphism (SNP). The common objective of imputation methods is to minimize the loss of information caused by experimental missing elements. In general, imputation of missing genotype data has used a major allele method, but this approach is not far from the objective of the imputation - minimizing the loss of information. This method generally produces high error rates of missing value estimation, since the characteristics of the genotype data are not considered over the structure of given genotype data. In our methods, we use the linkage disequilibrium and haplotype information for the missing SNP genotype. As a result, we provide the results of the comparative evaluation of our methods and major allele imputation method according to the various randomized missing rates.     

Impact of imputation of missing values on classification error for discrete data

Numerous industrial and research databases include missing values. It is not uncommon to encounter databases that have up to a half of the entries missing, making it very difficult to mine them using data analysis methods that can work only with complete data. A common way of dealing with this problem is to impute (fill-in) the missing values. This paper evaluates how the choice of different imputation methods affects the performance of classifiers that are subsequently used with the imputed data. The experiments here focus on discrete data. This paper studies the effect of missing data imputation using five single imputation methods (a mean method, a Hot deck method, a Na?¨ve-Bayes method, and the latter two methods with a recently proposed imputation framework) and one multiple imputation method (a polytomous regression based method) on classification accuracy for six popular classifiers (RIPPER, C4.5, K-nearest-neighbor, support vector machine with polynomial and RBF kernels, and Na?¨ve-Bayes) on 15 datasets. This experimental study shows that imputation with the tested methods on average improves classification accuracy when compared to classification without imputation. Although the results show that there is no universally best imputation method, Na?¨ve-Bayes imputation is shown to give the best results for the RIPPER classifier for datasets with high amount (i.e., 40% and 50%) of missing data, polytomous regression imputation is shown to be the best for support vector machine classifier with polynomial kernel, and the application of the imputation framework is shown to be superior for the support vector machine with RBF kernel and K-nearest-neighbor. The analysis of the quality of the imputation with respect to varying amounts of missing data (i.e., between 5% and 50%) shows that all imputation methods, except for the mean imputation, improve classification error for data with more than 10% of missing data. Finally, some classifiers such as C4.5 and Na?¨ve-Bayes were found to be missing data resistant, i.e., they can produce accurate classification in the presence of missing data, while other classifiers such as K-nearest-neighbor, SVMs and RIPPER benefit from the imputation.     

Observer design and model augmentation for bias compensation with a truck engine application

A systematic design method for reducing bias in observers is developed. The method utilizes an observable default model of the system together with measurement data from the real system and estimates a model augmentation. The augmented model is then used to design an observer which reduces the estimation bias compared to an observer based on the default model. Three main results are a characterization of possible augmentations from observability perspectives, a parameterization of the augmentations from the method, and a robustness analysis of the proposed augmentation estimation method. The method is applied to a truck engine where the resulting augmented observer reduces the estimation bias by 50% in a European transient cycle.     

Premium and protection of a response estimation procedure for two-way tables when outliers occur

The Q_K statistic is employed to detect spurious observations in two-way tables, and the protection afforded by an estimation rule that omits response values detected as outliers is then examined. The main conclusions are that small and moderate size discrepancies are difficult to detect and that the protection afforded is surprisingly insensitive to the test level used for the Q_K statistic.     

INSAT-3D and MODIS retrieved sea surface temperature validation and assessment over waters surrounding the Indian subcontinent

The INSAT-3D imager (4 km) and Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on-board Aqua and Terra space-platforms level-2 (1 km) sea surface temperature (SST_skin) product accuracy has been analysed over waters surrounding the Indian subcontinent by indirect comparison method using collocated bulk in-situ measurements (SST_depth) for 3 years (October 2013–October 2016). Statistical results show that root mean square error of all the three satellites is in range of around 0.60–0.70°C. Retrieval error is found to be slightly more in case of validation against iQuam data set. INSAT-3D is showing more underestimation with bias ranging from about ?0.16°C to ?0.20°C than MODIS sensor having bias in range of about 0.06°C to ?0.12°C. All the three missions are slightly underestimating over open-ocean with bias ranging in 0–0.17°C. INSAT-3D is significantly underestimating in-situ observations over the Arabian Sea (approximate bias = 0.27°C). Seasonal validation analysis reveals relatively high retrieval error during monsoon season than pre-monsoon and post-monsoon seasons. MODIS sensor is showing significant underestimation during monsoon with bias ranging from approximately ?0.29°C to ?0.58°C. Overall, all the three missions are performing similarly well over the study area.     

Improvements in land surface temperature retrieval based on atmospheric water vapour content and atmospheric temperature

Three methods are currently used to retrieve land surface temperatures (LSTs) from thermal infrared data supplied by the Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) sensors: the radiative transfer equation, mono-window, and generalized single-channel algorithms. Most retrieval results obtained using these three methods have an average error of more than 1 K. But if the regional mean atmospheric water vapour content and temperature are supplied by in situ radiosounding observations, the mono-window algorithm is able to provide better results, with a mean error of 0.5 K. However, there are no in situ radiosounding data for most regions. This article provides an improved method to retrieve LST from Landsat TM and ETM+ data using atmospheric water vapour content and atmospheric temperature, which can be obtained from remote-sensing data. The atmospheric water vapour content at the pixel scale was first calculated from Moderate Resolution Imaging Spectroradiometer (MODIS) data. The emissivities of various land covers and uses were then defined by Landsat TM or ETM+ data. In addition, the temperature–vegetation index method was applied to map area-wide instantaneous near-surface air temperatures. The parameters of mean atmospheric water vapour content and temperature and land surface emissivity were finally inputted to the mono-window algorithm to improve the LST retrieval precision. Our results indicate that this improved mono-window algorithm gave a significantly better retrieval of the estimated LST than that using the standard mono-window algorithm, not only in dry and elevated mountain regions but also in humid regions, as shown by the bias, standard deviation (σ), and root mean square deviation (RMSD). In Madoi County, the improved mono-window algorithm validated against the LST values measured in situ produced a bias and RMSD of –0.63 K and 0.91 K, respectively, compared with the mono-window algorithm’s bias and RMSD of –1.08 K and 1.27 K. Validated against the radiance-based method, the improved algorithm shows bias and RMSD values of –1.08 K and 1.27 K, respectively, compared with the initial algorithm’s bias and RMSD –1.65 K and 1.75 K. Additionally, the improved mono-window algorithm also appeared to be more accurate than the mono-window algorithm, with lower error values when validated against in situ measurement and the radiance-based method in the validation area in Zhangye City, Gansu Province, China. Remarkable LST accuracy improvements are shown by the improved mono-window algorithm, with better agreement not only with the in situ measurements but also with the simulated LSTs in the two validation areas, indicating the soundness and suitability of this method.     

An exact permutation method for testing any effect in balanced and unbalanced fixed effect ANOVA

The ANOVA method and permutation tests, two heritages of Fisher, have been extensively studied. Several permutation strategies have been proposed by others to obtain a distribution-free test for factors in a fixed effect ANOVA (i.e., single error term ANOVA). The resulting tests are either approximate or exact. However, there exists no universal exact permutation test which can be applied to an arbitrary design to test a desired factor. An exact permutation strategy applicable to fixed effect analysis of variance is presented. The proposed method can be used to test any factor, even in the presence of higher-order interactions. In addition, the method has the advantage of being applicable in unbalanced designs (all-cell-filled), which is a very common situation in practice, and it is the first method with this capability. Simulation studies show that the proposed method has an actual level which stays remarkably close to the nominal level, and its power is always competitive. This is the case even with very small datasets, strongly unbalanced designs and non-Gaussian errors. No other competitor show such an enviable behavior.     

面向推荐系统中有偏和无偏一元反馈建模的三任务变分自编码器

一元反馈建模在推荐系统中的应用非常广泛,例如点击预测和购买预测等。然而,推荐系统作为一个闭环的反馈系统,在用户与系统的交互过程中可能存在着多种偏置问题,例如位置偏置、流行偏置等,进而导致用户的反馈数据存在有偏性。现有的大部分推荐模型都只基于这样的有偏数据进行构建,忽略了偏置的影响,进而导致推荐结果是次优的。目前已有的偏置消除方法大致可以分为基于反事实学习的方法、基于启发式的方法和基于无偏数据增强的方法。其中,基于无偏数据增强的方法通常被认为在稳定性和准确性方面表现较好。
本文重点研究了推荐系统中一元反馈的偏置问题,通过引入由一种特定策略收集的无偏数据,结合有偏数据进行联合建模,从而得到更准确和无偏的推荐模型。具体而言,本文从多任务学习的角度对问题进行建模,将有偏数据、无偏数据以及它们的并集当作三种相互关联的信号,并设计了三个不同但相关的学习任务。变分自编码器是目前最先进的一元反馈建模方法之一,有着独特的建模方式,从而使得它在很多问题中取得了优越的推荐效果。本文基于变分自编码器提出了一种新的推荐模型,即三任务变分自编码器（Tri-VAE）。该模型包含三个变分自编码器,分别对三种信号进行重构。三个变分自编码器之间共享同一个编码器和同一个解码器。此外,本文还设计了特征校正模块和标签增强模块以加强任务之间的关联。其中,特征校正模块用来校正用户的潜在特征,得到更无偏的潜在特征,进而从潜在特征的角度缓解偏置的影响。标签增强模块用于生成可靠性较高的伪标签并加以利用,进而更有效地利用无偏数据中的信息。
在Yahoo!R3和Coat Shopping两个公开数据集上的实验结果表明,所提出的模型相比于目前最新的基线模型在绝大多数情况下取得了显著的效果提升。为了进一步研究所提出的模型,本文进行了消融实验、超参数敏感性分析和收敛性分析,还对特征校正模块的有效性进行了探讨。     

The CM-SAF operational scheme for the satellite based retrieval of solar surface irradiance — A LUT based eigenvector hybrid approach

The radiation budget at the earth surface is an essential climate variable for climate monitoring and analysis as well as for verification of climate model output and reanalysis data. Accurate solar surface irradiance data is a prerequisite for an accurate estimation of the radiation budget and for an efficient planning and operation of solar energy systems.This paper describes a new approach for the retrieval of the solar surface irradiance from satellite data. The method is based on radiative transfer modelling and enables the use of extended information about the atmospheric state. Accurate analysis of the interaction between the atmosphere, surface albedo, transmission and the top of atmosphere albedo has been the basis for the new method, characterised by a combination of parameterisations and “eigenvector” look-up tables. The method is characterised by a high computing performance combined with a high accuracy. The performed validation shows that the mean absolute deviation is of the same magnitude as the confidence level of the BSRN (Baseline Surface Radiation Measurement) ground based measurements and significant lower as the CM-SAF (Climate Monitoring Satellite Application Facility) target accuracy of 10 W/m². The mean absolute difference between monthly means of ground measurements and satellite based solar surface irradiance is 5 W/m² with a mean bias deviation of − 1 W/m² and a RMSD (Root Mean Square Deviation) of 5.4 W/m² for the investigated European sites. The results for the investigated African sites obtained by comparing instantaneous values are also encouraging. The mean absolute difference is with 2.8% even lower as for the European sites being 3.9%, but the mean bias deviation is with − 1.1% slightly higher as for the European sites, being 0.8%. Validation results over the ocean in the Mediterranean Sea using shipboard data complete the validation. The mean bias is − 3.6 W/m² and 2.3% respectively. The slightly higher mean bias deviation over ocean is at least partly resulting from inherent differences due to the movement of the ship (shadowing, allocation of satellite pixel). The validation results demonstrate that the high accuracy of the surface solar irradiance is given in different climate regions. The discussed method has also the potential to improve the treatment of radiation processes in climate and Numerical Weather Prediction (NWP) models, because of the high accuracy combined with a high computing speed.     

On the performance of bias-reduction techniques for variance estimation in approximate Bayesian bootstrap imputation

Multiply imputed data sets can be created with the approximate Bayesian bootstrap (ABB) approach under the assumption of ignorable nonresponse. The theoretical development and inferential validity are predicated upon asymptotic properties; and biases are known to occur in small-to-moderate samples. There have been attempts to reduce the finite-sample bias for the multiple imputation variance estimator. In this note, we present an empirical study for evaluating the comparative performance of the two proposed bias-correction techniques and their impact on precision. The results suggest that to varying degrees, bias improvements are outweighed by efficiency losses for the variance estimator. We argue that the original ABB has better small-sample properties than the modified versions in terms of the integrated behavior of accuracy and precision, as measured by the root mean-square error.     

Comparison of Bayesian survival analysis and Cox regression analysis in simulated and breast cancer data sets

We aimed to compare the performance of Cox regression analysis (CRA) and Bayesian survival analysis (BSA) by using simulations and breast cancer data.Simulation study was carried out with two different algorithms that were informative and noninformative priors. Moreover, in a real data set application, breast cancer data set related to disease-free survival (DFS) that was obtained from 423 breast cancer patients diagnosed between 1998 and 2007 was used.In the simulation application, it was observed that BSA with noninformative priors and CRA methods showed similar performances in point of convergence to simulation parameter. In the informative priors’ simulation application, BSA with proper informative prior showed a good performance with too little bias. It was found out that the bias of BSA increased while priors were becoming distant from reliability in all sample sizes. In addition, BSA obtained predictions with more little bias and standard error than the CRA in both small and big samples in the light of proper priors.In the breast cancer data set, age, tumor size, hormonal therapy, and axillary nodal status were found statistically significant prognostic factors for DFS in stepwise CRA and BSA with informative and noninformative priors. Furthermore, standard errors of predictions in BSA with informative priors were observed slightly.As a result, BSA showed better performance than CRA, when subjective data analysis was performed by considering expert opinions and historical knowledge about parameters. Consequently, BSA should be preferred in existence of reliable informative priors, in the contrast cases, CRA should be preferred.     

Estimation of classifier performance

An expression for expected classifier performance previously derived by the authors (ibid., vol.11, no.8, p.873-855, Aug. 1989) is applied to a variety of error estimation methods and a unified and comprehensive approach to the analysis of classifier performance is presented. After the error expression is introduced, it is applied to three cases: (1) a given classifier and a finite test set; (2) given test distributions a finite design set; and (3) finite and independent design and test sets. For all cases, the expected values and variances of the classifier errors are presented. Although the study of Case 1 does not produce any new results, it is important to confirm that the proposed approach produces the known results, and also to show how these results are modified when the design set becomes finite, as in Cases 2 and 3. The error expression is used to compute the bias between the leave-one-out and resubstitution errors for quadratic classifiers. The effect of outliers in design samples on the classification error is discussed. Finally, the theoretical analysis of the bootstrap method is presented for quadratic classifiers     

A parametric bootstrap approach for ANOVA with unequal variances: Fixed and random models

This article is about testing the equality of several normal means when the variances are unknown and arbitrary, i.e., the set up of the one-way ANOVA. Even though several tests are available in the literature, none of them perform well in terms of Type I error probability under various sample size and parameter combinations. In fact, Type I errors can be highly inflated for some of the commonly used tests; a serious issue that appears to have been overlooked. We propose a parametric bootstrap (PB) approach and compare it with three existing location-scale invariant tests—the Welch test, the James test and the generalized F (GF) test. The Type I error rates and powers of the tests are evaluated using Monte Carlo simulation. Our studies show that the PB test is the best among the four tests with respect to Type I error rates. The PB test performs very satisfactorily even for small samples while the Welch test and the GF test exhibit poor Type I error properties when the sample sizes are small and/or the number of means to be compared is moderate to large. The James test performs better than the Welch test and the GF test. It is also noted that the same tests can be used to test the significance of the random effect variance component in a one-way random model under unequal error variances. Such models are widely used to analyze data from inter-laboratory studies. The methods are illustrated using some examples.     

A parametric bootstrap approach for ANOVA with unequal variances: Fixed and random models

This article is about testing the equality of several normal means when the variances are unknown and arbitrary, i.e., the set up of the one-way ANOVA. Even though several tests are available in the literature, none of them perform well in terms of Type I error probability under various sample size and parameter combinations. In fact, Type I errors can be highly inflated for some of the commonly used tests; a serious issue that appears to have been overlooked. We propose a parametric bootstrap (PB) approach and compare it with three existing location-scale invariant tests—the Welch test, the James test and the generalized F (GF) test. The Type I error rates and powers of the tests are evaluated using Monte Carlo simulation. Our studies show that the PB test is the best among the four tests with respect to Type I error rates. The PB test performs very satisfactorily even for small samples while the Welch test and the GF test exhibit poor Type I error properties when the sample sizes are small and/or the number of means to be compared is moderate to large. The James test performs better than the Welch test and the GF test. It is also noted that the same tests can be used to test the significance of the random effect variance component in a one-way random model under unequal error variances. Such models are widely used to analyze data from inter-laboratory studies. The methods are illustrated using some examples.     

An analysis of three-level orthogonal saturated designs

Although three-level factorial designs with quantitative factors are not the most efficient way to fit a second-order polynomial model, they often find some application, when the factors are qualitative. The three-level orthogonal designs with qualitative factors are frequently used, e.g., in agriculture, in clinical trials and in parameter designs. It is practically unavoidable that, because of the limitation of experimental materials or time-related constraint, we often have to keep the number of experiments as small as possible and to consider the effects of many factors and interactions simultaneously so that most of such designs are saturated or nearly saturated. An experimental design is said to be saturated, if all degrees of freedom are consumed by the estimation of parameters in modelling mean response. The difficulty of analyzing orthogonal saturated designs is that there are no degrees of freedom left to estimate the error variance so that the ordinary ANOVA is no longer available. In this paper, we present a new formal test, which is based on mean squares, for analyzing three-level orthogonal saturated designs. This proposed method is compared via simulation with several mean squares based methods published in the literature. The results show that the new method is more powerful in terms of empirical power of the test. Critical values used in the proposed procedure for some three-level saturated designs are tabulated. Industrial examples are also included for illustrations.     

Augmented instrumental variable method for position and heading estimation with RDOA measurements

In this paper, we proposed a position and heading estimation algorithm using only range difference of arrival (RDOA) measurements. Based on RDOA measurements, an uncertain linear measurement model is derived and both position and heading are estimated with the instrumental variable (IV) method which can show unbiased estimation results for the uncertainty of the model. In addition, to remove the unknown bias included in the measurement model error, we augment the bias to the state vector of the model. Since the proposition inherits the characteristic of the IV method, it does not need the stochastic information of the RDOA measurement excepting the assumption that the RDOA measurement noise is zero mean and white, and the zero mean error performance can be guaranteed when variances of RDOA measurement noises are identical. Through simulations, the performance of the proposed algorithm is verified at various positions and headings in the sensor network and compared with the robust least squares method which shows a zero mean error performance under the assumption that the stochastic information is known exactly.     

MIMO two-way relay channel with superposition coding and imperfect channel estimation

This paper deals with the superposition coding (SPC) scheme in multiple-input multiple-output two-way relay channels subject to imperfect channel estimation. In this scenario, two multiple antenna terminals, which are unable to communicate directly, exchange information with each other via a multiple antenna relay. We determine the impact of the channel estimation error degradation on the achievable rate region for two main SPC techniques: (a) SPC without channel state information (CSI) at the users, (b) SPC with an imperfect CSI at the users where a waterfilling power allocation is employed. We demonstrate that imperfect CSI significantly improves the achievable rate at low signal-to-noise ratios (SNRs) while it becomes less critical at high SNRs. In addition, a SPC power allocation technique that incorporates the average channel statistics and does not require any instantaneous CSI is also investigated. We show how the available power is split between the two bi-directional (superimposed) data flows in order to maximize the system performance and to support fairness as well as to maximize the achievable sum-rate.     

Bias-free identification of a linear model-predictive steering controller from measured driver steering behavior

Recent developments in modeling driver steering control with preview are reviewed. While some validation with experimental data has been presented, the rigorous application of formal system identification methods has not yet been attempted. This paper describes a steering controller based on linear model-predictive control. An indirect identification method that minimizes steering angle prediction error is developed. Special attention is given to filtering the prediction error so as to avoid identification bias that arises from the closed-loop operation of the driver-vehicle system. The identification procedure is applied to data collected from 14 test drivers performing double lane change maneuvers in an instrumented vehicle. It is found that the identification procedure successfully finds parameter values for the model that give small prediction errors. The procedure is also able to distinguish between the different steering strategies adopted by the test drivers.