An incorporative statistic and neural approach for crop yield modelling and forecasting

An incorporative framework is proposed in this study for crop yield modelling and forecasting. It is a complementary approach to traditional time series analysis on modelling and forecasting by treating crop yield and associated factors as a non-temporal collection. Statistics are used to identify the highly related factor(s) among many associates to crop yield and then play a key role in data cleaning and a supporting role in data expansion, if necessary, for neural network training and testing. Wheat yield and associated plantation area, rainfall and temperature in Queensland of Australia over 100 years are used to test this incorporative approach. The results show that well-trained multilayer perceptron models can simulate the wheat production through given plantation areas with a mean absolute error (MAE) of ~2%, whereas the third-order polynomial correlation returns an MAE of ~20%. However, statistical analysis plays a key role in identifying the most related factor, detecting outliers, determining the general trend of wheat yield with respect to plantation area and supporting data expansion for neural network training and testing. The combination of these two methods provides both meaningful qualitative and accurate quantitative data analysis and forecasting. This incorporative approach can also be useful in data modelling and forecasting in other applications due to its generic nature.     

A Framework for Robust Subspace Learning

Many computer vision, signal processing and statistical problems can be posed as problems of learning low dimensional linear or multi-linear models. These models have been widely used for the representation of shape, appearance, motion, etc., in computer vision applications. Methods for learning linear models can be seen as a special case of subspace fitting. One draw-back of previous learning methods is that they are based on least squares estimation techniques and hence fail to account for outliers which are common in realistic training sets. We review previous approaches for making linear learning methods robust to outliers and present a new method that uses an intra-sample outlier process to account for pixel outliers. We develop the theory of Robust Subspace Learning (RSL) for linear models within a continuous optimization framework based on robust M-estimation. The framework applies to a variety of linear learning problems in computer vision including eigen-analysis and structure from motion. Several synthetic and natural examples are used to develop and illustrate the theory and applications of robust subspace learning in computer vision.     

基于最大相关熵准则的2维主成分分析

目的 本文针对基于最小均方差准则的主成分分析算法(如2DPCA-L₂(two-dimensional PCA with L₂-norm)算法和2DPCA-L₁(two-dimensional PCA with L₁-norm)算法)对外点敏感、识别率低的问题,结合信息论中的最大相关熵准则,提出了一种基于最大相关熵准则的2DPCA(2DPCA-MCC)。方法 2DPCA-MCC算法采用最大相关熵表示目标函数,通过半二次优化技术解决相关熵问题,降低了外点在目标函数评价中的贡献,从而提高了算法的鲁棒性和识别精度。结果 通过对比2DPCA-MCC算法和2DPCA-L₂、2DPCA-L₁在ORL人脸数据库上的识别效果,表明了2DPCA-MCC算法的识别率比2维主成分分析算法的识别率最低提高了近10%,最高提高了近30%。结论 提出了一种基于最大相关熵的2DPCA算法,通过半二次优化技术解决非线性优化问题,实验结果表明,本算法能够较好地解决外点问题,显著提高识别精度,适用于解决人脸识别中的外点问题。     

Towards a framework for multiple artificial neural network topologies validation by means of statistics

Artificial neural networks (ANNs) are flexible computing tools that have been applied to a wide range of domains with a notable level of accuracy. However, there are multiple choices of ANNs classifiers in the literature that produce dissimilar results. As a consequence of this, the selection of this classifier is crucial for the overall performance of the system. In this work, an integral framework is proposed for the optimization of different ANN classifiers based on statistical hypothesis testing. The framework is tested in a real ballistic scenario. The new quality measures introduced, based on the Student t‐test, and employed throughout the framework, ensure the validity of results from a statistical standpoint; they reduce the appearance of experimental errors or the appearance of possible randomness. Results show the relevance of this framework, proving that its application improves the performance and efficiency of multiple classifiers.     

Preliminary Data Analysis Methods in Software Estimation

Software is quite often expensive to develop and can become a major cost factor in corporate information systems budgets. With the variability of software characteristics and the continual emergence of new technologies the accurate prediction of software development costs is a critical problem within the project management context. In order to address this issue a large number of software cost prediction models have been proposed. Each model succeeds to some extent but they all encounter the same problem, i.e., the inconsistency and inadequacy of the historical data sets. Often a preliminary data analysis has not been performed and it is possible for the data to contain non-dominated or confounded variables. Moreover, some of the project attributes or their values are inappropriately out of date, for example the type of computer used for project development in the COCOMO 81 (Boehm, 1981) data set. This paper proposes a framework composed of a set of clearly identified steps that should be performed before a data set is used within a cost estimation model. This framework is based closely on a paradigm proposed by Maxwell (2002). Briefly, the framework applies a set of statistical approaches, that includes correlation coefficient analysis, Analysis of Variance and Chi-Square test, etc., to the data set in order to remove outliers and identify dominant variables. To ground the framework within a practical context the procedure is used to analyze the ISBSG (International Software Benchmarking Standards Group data—Release 8) data set. This is a frequently used accessible data collection containing information for 2,008 software projects. As a consequence of this analysis, 6 explanatory variables are extracted and evaluated.     

基于鲁棒相关向量机的转炉炼钢终点预报模型

针对传统相关向量机在训练过程中易受异常点影响的问题,提出了一种鲁棒相关向量机模型,并将其应用于转炉炼钢终点碳含量和温度的预报.通过为每一个训练样本设定独立的噪声方差系数,并使其在训练过程中随模型预测误差的增大而逐渐减小来降低异常点的影响,同时依据贝叶斯证据框架给出了模型超参数的迭代计算公式,进行参数的优化.使用标准测试数据和转炉炼钢实际生产数据进行仿真,结果表明本文模型具有较好的预报精度和鲁棒性.     

A variance shift model for detection of outliers in the linear mixed model

A variance shift outlier model (VSOM), previously used for detecting outliers in the linear model, is extended to the variance components model. This VSOM accommodates outliers as observations with inflated variance, with the status of the ith observation as an outlier indicated by the size of the associated shift in the variance. Likelihood ratio and score test statistics are assessed as objective measures for determining whether the ith observation has inflated variance and is therefore an outlier. It is shown that standard asymptotic distributions do not apply to these tests for a VSOM, and a modified distribution is proposed. A parametric bootstrap procedure is proposed to account for multiple testing. The VSOM framework is extended to account for outliers in random effects and is shown to have an advantage over case-deletion approaches. A simulation study is presented to verify the performance of the proposed tests. Challenges associated with computation and extensions of the VSOM to the general linear mixed model with correlated errors are discussed.     

A practical model‐based statistical approach for generating functional test cases: application in the automotive industry

With the growing complexity of industrial software applications, industrials are looking for efficient and practical methods to validate the software. This paper develops a model‐based statistical testing approach that automatically generates online and offline test cases for embedded software. It discusses an integrated framework that combines solutions for three major software testing research questions: (i) how to select test inputs; (ii) how to predict the expected results of a test; and (iii) when to stop testing software. The automatic selection of test inputs is based on a stochastic test model that accounts for the main particularity of embedded software: time sensitivity. Software test practitioners may design one or more test models when they generate random, user‐oriented, or fault‐oriented test inputs. A formal framework integrating existing and appropriate specification techniques was developed for the design of automated test oracles (executable software specifications) and the formal measurement of functional coverage. The decision to stop testing software is based on both test coverage objectives and cost constraints. This approach was tested on two representative case studies from the automotive industry. The experiment was performed at unit testing level in a simulated environment on a host personal computer (automatic test execution). The two software functionalities tested had previously been unit tested and validated using the test design approach conventionally used in the industry. Applying the proposed model‐based statistical testing approach to these two case studies, we obtained significant improvements in performing functional unit testing in a real and complex industrial context: more bugs were detected earlier and in a shorter time. Copyright © 2012 John Wiley & Sons, Ltd.     

Results analysis for trust constrained real-time optimization

In real-time optimization (RTO), results analysis is used to ensure that RTO predictions can be implemented and are not the result of the unnecessary variance transmission around the RTO loop. Miletic and Marlin [2] proposed a statistical framework for analyzing RTO results; however, their method cannot effectively deal with inequality constraints. Many industrial RTO implementations include bounds on the changes that the RTO system can make to the process operation (i.e. trust-region constraints). Such trust-region constraints can seriously degrade the performance of existing results analysis methods. In this paper, a results analysis procedure is proposed that incorporates statistical testing on both the primal and dual variables of the optimization problem to effectively analyze steady-state RTO results in the presence of trust-region constraints. The proposed method is illustrated using two small case studies, one of which is the same Williams and Otto [11] reactor example used in [2].     

考虑时延偏差的数字电路时延测试综述

先进集成电路工艺下,时延测试是数字电路测试的一项重要内容。各种时延偏差来源如小时延缺陷、工艺偏差、 串扰、电源噪声、老化效应等,影响着电路的额定时钟频率,是时延测试中需要考虑的因素。文章在介绍电路时延偏差 问题的各种来源的基础上,给出了针对不同的时延偏差问题所涉及的分析、建模、测试生成与电路设计等关键技术。进 一步介绍了中国科学院计算技术研究所近年来在考虑时延偏差的数字电路时延测试方面所做的研究工作,包括：考虑串 扰/电源噪声的时延测试、基于统计定时分析的测试通路选择、片上时延测量、超速测试、测试优化、在线时序检测等方 面。文章最后对数字电路时延测试技术的发展趋势进行了总结。     

Using combinatorial optimization in model-based trimmed clustering with cardinality constraints

Statistical clustering criteria with free scale parameters and unknown cluster sizes are inclined to create small, spurious clusters. To mitigate this tendency a statistical model for cardinality-constrained clustering of data with gross outliers is established, its maximum likelihood and maximum a posteriori clustering criteria are derived, and their consistency and robustness are analyzed. The criteria lead to constrained optimization problems that can be solved by using iterative, alternating trimming algorithms of k-means type. Each step in the algorithms requires the solution of a λ-assignment problem known from combinatorial optimization. The method allows one to estimate the numbers of clusters and outliers. It is illustrated with a synthetic data set and a real one.     

Extracting lines of curvature from noisy point clouds

We present a robust framework for extracting lines of curvature from point clouds. First, we show a novel approach to denoising the input point cloud using robust statistical estimates of surface normal and curvature which automatically rejects outliers and corrects points by energy minimization. Then the lines of curvature are constructed on the point cloud with controllable density. Our approach is applicable to surfaces of arbitrary genus, with or without boundaries, and is statistically robust to noise and outliers while preserving sharp surface features. We show our approach to be effective over a range of synthetic and real-world input datasets with varying amounts of noise and outliers. The extraction of curvature information can benefit many applications in CAD, computer vision and graphics for point cloud shape analysis, recognition and segmentation. Here, we show the possibility of using the lines of curvature for feature-preserving mesh construction directly from noisy point clouds.     

Exact computation of the null distribution of a test for multiple outliers in an exponential sample

Through a systematic application of the recursion of Huffer [Huffer, F., 1988. Divided differences and the joint distribution of linear combinations of spacings. Journal of Applied Probability 25, 346-354], we present an algorithm for evaluating the exact null distribution of the test statistic proposed by Kimber [Kimber, A.C., 1982. Tests for many outliers in an exponential sample. Applied Statistics 31, 263-271] for the testing of up to k upper outliers for discordancy in exponential samples. This method presents another way of obtaining the exact null distribution of the test statistic without first obtaining their joint density. The advantage of this approach is in its generality and also in its ease of use. In addition, it can provide critical values for the test statistic that are accurate to any required degree of precision. Similar results for the sequential testing of up to k lower outliers for discordancy are also presented.     

Formal testing for separation assurance

In order to address the rapidly increasing load of air traffic operations, innovative algorithms and software systems must be developed for the next generation air traffic control. Extensive verification of such novel algorithms is key for their adoption by industry. Separation assurance algorithms aim at predicting if two aircraft will get closer to each other than a minimum safe distance; if loss of separation is predicted, they also propose a change of course for the aircraft to resolve this potential conflict. In this paper, we report on our work towards developing an advanced testing framework for separation assurance. Our framework supports automated test case generation and testing, and defines test oracles that capture algorithm requirements. We discuss three different approaches to test-case generation, their application to a separation assurance prototype, and their respective strengths and weaknesses. We also present an approach for statistical analysis of the large numbers of test results obtained from our framework.     

Consistent Depth Maps Recovery from a Video Sequence

This paper presents a novel method for recovering consistent depth maps from a video sequence. We propose a bundle optimization framework to address the major difficulties in stereo reconstruction, such as dealing with image noise, occlusions, and outliers. Different from the typical multi-view stereo methods, our approach not only imposes the photo-consistency constraint, but also explicitly associates the geometric coherence with multiple frames in a statistical way. It thus can naturally maintain the temporal coherence of the recovered dense depth maps without over-smoothing. To make the inference tractable, we introduce an iterative optimization scheme by first initializing the disparity maps using a segmentation prior and then refining the disparities by means of bundle optimization. Instead of defining the visibility parameters, our method implicitly models the reconstruction noise as well as the probabilistic visibility. After bundle optimization, we introduce an efficient space-time fusion algorithm to further reduce the reconstruction noise. Our automatic depth recovery is evaluated using a variety of challenging video examples.     

自动化单元测试框架NUnit的改进设计与实现

软件测试对确保软件质量有着不可替代的作用。自动化测试框架有效提高了测试效率,自动化测试框架的成熟是软件测试走向标准化的必经之路。文章改进了自动化单元测试框架NUnit,使测试代码和测试数据分离,解决了使用NUnit测试时测试代码存在大量冗余的问题。在改进的NUnit框架中,相似的测试用例只需测试人员编写一次,框架将自动生成其他测试用例。     

Principle component analysis: Robust versions

Modern problems of optimization, estimation, signal and image processing, pattern recognition, etc., deal with huge-dimensional data; this necessitates elaboration of efficient methods of processing such data. The idea of building low-dimensional approximations to huge data arrays is in the heart of the modern data analysis.One of the most appealing methods of compact data representation is the statistical method referred to as the principal component analysis; however, it is sensitive to uncertainties in the available data and to the presence of outliers. In this paper, robust versions of the principle component analysis approach are proposed along with numerical methods for their implementation.     

A new linear discriminant analysis algorithm based on L1-norm maximization and locality preserving projection

Generic L2-norm-based linear discriminant analysis (LDA) is sensitive to outliers and only captures global structure information of sample points. In this paper, a new LDA-based feature extraction algorithm is proposed to integrate both global and local structure information via a unified L1-norm optimization framework. Unlike generic L2-norm-based LDA, the proposed algorithm explicitly incorporates the local structure information of sample points and is robust to outliers. It overcomes the problem of the singularity of within-class scatter matrix as well. Experiments on several popular datasets demonstrate the effectiveness of the proposed algorithm.     

A framework of model validation and virtual product qualification with limited experimental data based on statistical inference

Virtual testing is a recent engineering development trend to design, evaluate, and test new engineered products. This research proposes a framework of virtual testing based on statistical inference for new product development comprising of three successive steps: (i) statistical model calibration, (ii) hypothesis test for validity check and (iii) virtual qualification. Statistical model calibration first improves the predictive capability of a computational model in a calibration domain. Next, the hypothesis test is performed with limited observed data to see if a calibrated model is sufficiently predictive for virtual testing of a new product design. An area metric and the u-pooling method are employed for the hypothesis test to measure the degree of mismatch between predicted and observed results while considering statistical uncertainty in the area metric due to the lack of experimental data. Once the calibrated model becomes valid, the virtual qualification process can be executed with a qualified model for new product developments. The qualification process builds a design decision matrix to aid in rational decision-making for product design alternatives. The effectiveness of the proposed framework is demonstrated through the case study of a tire tread block.     

An adaptive and trustworthy software testing framework on the grid

Grid computing, which is characterized by large-scale sharing and collaboration of dynamic distributed resources has quickly become a mainstream technology in distributed computing and is changing the traditional way of software development. In this article, we present a grid-based software testing framework for unit and integration test, which takes advantage of the large-scale and cost-efficient computational grid resources to establish a testbed for supporting automated software test in complex software applications. Within this software testing framework, a dynamic bag-of-tasks model using swarm intelligence is developed to adaptively schedule unit test cases. Various high-confidence computing mechanisms, such as redundancy, intermediate value checks, verification code injection, and consistency checks are employed to verify the correctness of each test case execution on the grid. Grid workflow is used to coordinate various test units for integration test. Overall, we expect that the grid-based software testing framework can provide efficient and trustworthy services to significantly accelerate the testing process with large-scale software testing.