Search

The paper deals with a robust autonomous path-following functionality, in which the safe velocity and the motion profile of the vehicle with varying tyre-road contact must be guaranteed. The integration of Linear Parameter-Varying (LPV) control and the results of the machine learning-based analysis on the big data of autonomous vehicles is proposed. The integration is achieved in two steps. First, an estimation method on the adhesion coefficient through decision trees is presented. The result of the estimation is incorporated in the robust LPV control through a scheduling variable. During the control design, the error of the machine-learning algorithm is incorporated. Second, an optimisation method of the longitudinal velocity on a predicted horizon is proposed, which is aided with the machine learning-based reachability set approximation of the steering intervention. The effectiveness of the control strategy is illustrated through CarSim simulations. [ABSTRACT FROM AUTHOR]

Considering looming fatality and economic recession, effective policy making based on ongoing COVID-19 pandemic is an urgent and standing issue. Numerous issues for controlling infection have arisen from public discussion led by medical professionals. Yet understanding of these factors has been necessarily qualitative and control measures to correct unfavorable trends specific to an infection area have been lacking. The logical implement for control is a large scale stochastic model with countless parameters lacking robustness and requiring enormous data. This paper presents a remedy for this vexing problem by proposing an alternative approach. Machine learning has come to play a widely circulated role in the study of complex data in recent times. We demonstrate that when machine learning is employed together with the mechanistic framework of a mathematical model, there can be a considerably enhanced understanding of complex systems. A mathematical model describing the viral infection dynamics reveals two transmissibility parameters influenced by the management strategies in the area for the control of the current pandemic. Both parameters readily yield the peak infection rate and means for flattening the curve, which is correlated to different management strategies by employing machine learning, enabling comparison of different strategies and suggesting timely alterations. Treatment of population data with the model shows that restricted non-essential business closure, school closing and strictures on mass gathering influence the spread of infection. While a rational strategy for initiation of an economic reboot would call for a wider perspective of the local economics, the model can speculate on its timing based on the status of the infection as reflected by its potential for an unacceptably renewed viral onslaught. [ABSTRACT FROM AUTHOR]

Machine learning methods have gained increased popularity in biomedical research during the recent years. However, very few of them support the analysis of longitudinal data, where several samples are collected from an individual over time. Additionally, most of the available longitudinal machine learning methods assume that the measurements are aligned in time, which is often not the case in real data. Here, we introduce a robust longitudinal machine learning method, named likelihood contrasts (LC), which supports study designs with unaligned time points. Our LC method is a binary classifier, which uses linear mixed models for modelling and log-likelihood for decision making. To demonstrate the benefits of our approach, we compared it with existing methods in four simulated and three real data sets. In each simulated data set, LC was the most accurate method, while the real data sets further supported the robust performance of the method. LC is also computationally efficient and easy to use. [ABSTRACT FROM AUTHOR]

The lack of adequate measures is often an impediment to robust policy evaluation. We discuss three approaches to measurement and data usage that have the potential to improve the way we conduct impact evaluations. First, the creation of new measures, when no adequate ones are available. Second, the use of multiple measures when a single one is not appropriate. And third, the use of machine learning algorithms to evaluate and understand programme impacts. We motivate the relevance of each of the categories by providing examples where they have proved useful in the past. We discuss the challenges and risks involved in each strategy and conclude with an outline of promising directions for future work. [ABSTRACT FROM AUTHOR]

In the era of big data, it is increasingly common that large amount of data is generated across multiple distributed sites and cannot be gathered into a centralized site for further analysis, which invalidates the assumption of traditional clustering techniques based on centralized models. The major challenge is that these distributed datasets cannot be trivially merged due to issues such as privacy concerns, limited network bandwidth among sites and limited computational capacity of a single site. To tackle this challenge, we propose an efficient distributed clustering scheme using boundary information (DCUBI), which features good flexibility and scalability. The main procedure of DCUBI consists of three steps: local-global-local. Firstly, each local site extracts the boundary points from its own local data and applies traditional clustering on boundary points only. Secondly, labeled boundary points from each site are sent to the central site as local representatives where boundary and cluster fusion is conducted to form the global clustering model. Finally, the global boundary and cluster information is sent back to each local site for refined local clustering. To demonstrate the effectiveness of DCUBI, we plug the well-known DBSCAN algorithm into DCUBI and comprehensive experiments are conducted using datasets with different properties. Experiment results clearly verify the quality of clustering by DCUBI as well as its superior time efficiency when the volume of data in each site is large. Furthermore, other popular clustering techniques especially those with high time complexity such as spectral clustering and affinity propagation clustering are also plugged into DCUBI to demonstrate the flexibility of the proposed scheme. [ABSTRACT FROM AUTHOR]

Mixture of experts (MoE) is a popular framework in the fields of statistics and machine learning for modeling heterogeneity in data for regression, classification and clustering. MoE for continuous data are usually based on the normal distribution. However, it is known that for data with asymmetric behavior, heavy tails and atypical observations, the use of the normal distribution is unsuitable. We introduce a new robust non-normal mixture of experts modeling using the skew t distribution. The proposed skew t mixture of experts, named STMoE, handles these issues of the normal mixtures experts regarding possibly skewed, heavy-tailed and noisy data. We develop a dedicated expectation conditional maximization (ECM) algorithm to estimate the model parameters by monotonically maximizing the observed data log-likelihood. We describe how the presented model can be used in prediction and in model-based clustering of regression data. Numerical experiments carried out on simulated data show the effectiveness and the robustness of the proposed model in fitting non-linear regression functions as well as in model-based clustering. Then, the proposed model is applied to the real-world data of tone perception for musical data analysis, and the one of temperature anomalies for the analysis of climate change data. The obtained results confirm the usefulness of the model for practical data analysis applications. [ABSTRACT FROM AUTHOR]

One-class classification belongs to the one of the novel and very promising topics in contemporary machine learning. In recent years ensemble approaches have gained significant attention due to increasing robustness to unknown outliers and reducing the complexity of the learning process. In our previous works, we proposed a highly efficient one-class classifier ensemble, based on input data clustering and training weighted one-class classifiers on clustered subsets. However, the main drawback of this approach lied in difficult and time consuming selection of a number of competence areas which indirectly affects a number of members in the ensemble. In this paper, we investigate ten different methodologies for an automatic determination of the optimal number of competence areas for the proposed ensemble. They have roots in model selection for clustering, but can be also effectively applied to the classification task. In order to select the most useful technique, we investigate their performance in a number of one-class and multi-class problems. Numerous experimental results, backed-up with statistical testing, allows us to propose an efficient and fully automatic method for tuning the one-class clustering-based ensembles. [ABSTRACT FROM AUTHOR]

Due to the fact that limited amounts of labeled data are normally available in real-world, semi-supervised learning has become a popular option, where we expect to use unlabeled data information to improve the learning performance. However, how to use such unlabeled information to make the predicted labels more reliable remains to be a key for any successful learning. In this paper, we propose a semi-supervised learning framework via combination of semi-supervised clustering and semi-supervised classification. In our approach, the predicted labels are selected by both the constrained k -means and safe semi-supervised SVM (S4VMs) to improve the reliability of the predicted labels. Extensive evaluations on collection of benchmarks and real-world action recognition datasets show that the proposed technique outperforms the others. [ABSTRACT FROM AUTHOR]

We present a novel approach to learn binary classifiers when only positive and unlabeled instances are available (PU learning). This problem is routinely cast as a supervised task with label noise in the negative set. We use an ensemble of SVM models trained on bootstrap resamples of the training data for increased robustness against label noise. The approach can be considered in a bagging framework which provides an intuitive explanation for its mechanics in a semi-supervised setting. We compared our method to state-of-the-art approaches in simulations using multiple public benchmark data sets. The included benchmark comprises three settings with increasing label noise: (i) fully supervised, (ii) PU learning and (iii) PU learning with false positives. Our approach shows a marginal improvement over existing methods in the second setting and a significant improvement in the third. [ABSTRACT FROM AUTHOR]

The class imbalance problem has attracted a lot of attention from the data mining community recently, becoming a current trend in machine learning research. The Consolidated Tree Construction (CTC) algorithm was proposed as an algorithm to solve a classification problem involving a high degree of class imbalance without losing the explaining capacity, a desirable characteristic of single decision trees and rule sets. CTC works by resampling the training sample and building a tree from each subsample, in a similar manner to ensemble classifiers, but applying the ensemble process during the tree construction phase, resulting in a unique final tree. In the ECML/PKDD 2013 conference the term “Inner Ensembles” was coined to refer to such methodologies. In this paper we propose a resampling strategy for classification algorithms that use multiple subsamples. This strategy is based on the class distribution of the training sample to ensure a minimum representation of all classes when resampling. This strategy has been applied to CTC over different classification contexts. A robust classification algorithm should not just be able to rank in the top positions for certain classification problems but should be able to excel when faced with a broad range of problems. In this paper we establish the robustness of the CTC algorithm against a wide set of classification algorithms with explaining capacity. [ABSTRACT FROM AUTHOR]

The existing studies involving matrix or tensor completion problems are commonly under the nuclear norm penalization framework due to the computational efficiency of the resulting convex optimization problem. Folded-concave penalization methods have demonstrated surprising developments in sparse learning problems due to their nice practical and theoretical properties. To share the same light of folded-concave penalization methods, we propose a new tensor completion model via folded-concave penalty for estimating missing values in tensor data. Two typical folded-concave penalties, the minmax concave plus (MCP) penalty and the smoothly clipped absolute deviation (SCAD) penalty, are employed in the new model. To solve the resulting nonconvex optimization problem, we develop a local linear approximation augmented Lagrange multiplier (LLA-ALM) algorithm which combines a two-step LLA strategy to search a local optimum of the proposed model efficiently. Finally, we provide numerical experiments with phase transitions, synthetic data sets, real image and video data sets to exhibit the superiority of the proposed model over the nuclear norm penalization method in terms of the accuracy and robustness. [ABSTRACT FROM AUTHOR]

Nonnegative matrix factorization (NMF) under the separability assumption can provably be solved efficiently, even in the presence of noise, and has been shown to be a powerful technique in document classification and hyperspectral unmixing. This problem is referred to as near-separable NMF and requires that there exists a cone spanned by a small subset of the columns of the input nonnegative matrix approximately containing all columns. In this paper, we propose a preconditioning based on the minimum volume ellipsoid and semidefinite programming making the input matrix well-conditioned. This in turn can improve significantly the performance of near-separable NMF algorithms which is illustrated on the popular successive projection algorithm (SPA). The new preconditioned SPA is provably more robust to noise, and outperforms SPA on several synthetic data sets. We also show how an active-set method allows us to apply the preconditioning on large-scale real-world hyperspectral images. [ABSTRACT FROM AUTHOR]

In this paper, we tackle the problem of estimating the correct number of clusters from a new perspective. The proposed method probabilistically estimates the Average Central Error (ACE), which is the difference between the true cluster centers and their estimations. The novelty of this work is partly in estimating the unavailable ACE by using the available cluster compactness that is the difference between estimated centers and their members. The application of this approach is explored with K-means clustering. The proposed method denoted by Minimum ACE K-means (MACE-means) is shown to have unique advantages both with synthetic and real data. MACE-means clustering is applied to benchmark real world data sets from UCI machine learning repository and other synthesized clusters that represent a wide class of clustering scenarios. Our analysis confirms superiority of MACE-means over the state of the art clustering methods in robustness to the initialization error, accuracy in detecting the correct number of clusters, having less time complexity, and robustness to cluster overlapping. [ABSTRACT FROM AUTHOR]

Abstract: Matrices, or more generally, multi-way arrays (tensors) are common forms of data that are encountered in a wide range of real applications. How to classify this kind of data is an important research topic for both pattern recognition and machine learning. In this paper, by analyzing the relationship between two famous traditional classification approaches, i.e., SVM and STM, a novel tensor-based method, i.e., multiple rank multi-linear SVM (MRMLSVM), is proposed. Different from traditional vector-based and tensor based methods, multiple-rank left and right projecting vectors are employed to construct decision boundary and establish margin function. We reveal that the rank of transformation can be regarded as a tradeoff parameter to balance the capacity of learning and generalization in essence. We also proposed an effective approach to solve the proposed non-convex optimization problem. The convergence behavior, initialization, computational complexity and parameter determination problems are analyzed. Compared with vector-based classification methods, MRMLSVM achieves higher accuracy and has lower computational complexity. Compared with traditional supervised tensor-based methods, MRMLSVM performs better for matrix data classification. Promising experimental results on various kinds of data sets are provided to show the effectiveness of our method. [Copyright &y& Elsevier]

SUMMARY Cloud computing is attracting the attention of the scientific community. In this paper, we develop a new cloud-based computing system in the Windows Azure platform that allows users to use the Zeolite Structure Predictor (ZSP) model through a Web browser. The ZSP is a novel machine learning approach for classifying zeolite crystals according to their framework type. The ZSP can categorize entries from the Inorganic Crystal Structure Database into 41 framework types. The novel automated system permits a user to calculate the vector of descriptors used by ZSP and to apply the model using the Random Forest™ algorithm for classifying the input zeolite entries. The workflow presented here integrates executables in Fortran and Python for number crunching with packages such as Weka for data analytics and Jmol for Web-based atomistic visualization in an interactive compute system accessed through the Web. The compute system is robust and easy to use. Communities of scientists, engineers, and students knowledgeable in Windows-based computing should find this new workflow attractive and easy to be implemented in scientific scenarios in which the developer needs to combine heterogeneous components. Copyright © 2012 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

A key problem in visual tracking is how to effectively combine spatio-temporal visual information from throughout a video to accurately estimate the state of an object. We address this problem by incorporating Dempster–Shafer (DS) information fusion into the tracking approach. To implement this fusion task, the entire image sequence is partitioned into spatially and temporally adjacent subsequences. A support vector machine (SVM) classifier is trained for object/nonobject classification on each of these subsequences, the outputs of which act as separate data sources. To combine the discriminative information from these classifiers, we further present a spatio-temporal weighted DS (STWDS) scheme. In addition, temporally adjacent sources are likely to share discriminative information on object/nonobject classification. To use such information, an adaptive SVM learning scheme is designed to transfer discriminative information across sources. Finally, the corresponding DS belief function of the STWDS scheme is embedded into a Bayesian tracking model. Experimental results on challenging videos demonstrate the effectiveness and robustness of the proposed tracking approach. [ABSTRACT FROM PUBLISHER]

Abstract: The advent of high throughput technologies, in particular microarrays, for biological research has revived interest in clustering, resulting in a plethora of new clustering algorithms. However, model selection, i.e., the identification of the correct number of clusters in a dataset, has received relatively little attention. Indeed, although central for statistics, its difficulty is also well known. Fortunately, a few novel techniques for model selection, representing a sharp departure from previous ones in statistics, have been proposed and gained prominence for microarray data analysis. Among those, the stability-based methods are the most robust and best performing in terms of prediction, but the slowest in terms of time. It is very unfortunate that as fascinating and classic an area of statistics as model selection, with important practical applications, has received very little attention in terms of algorithmic design and engineering. In this paper, in order to partially fill this gap, we make the following contributions: (A) the first general algorithmic paradigm for stability-based methods for model selection; (B) reductions showing that all of the known methods in this class are an instance of the proposed paradigm; (C) a novel algorithmic paradigm for the class of stability-based methods for cluster validity, i.e., methods assessing how statistically significant is a given clustering solution; (D) a general algorithmic paradigm that describes heuristic and very effective speed-ups known in the literature for stability-based model selection methods. Since the performance evaluation of model selection algorithms is mainly experimental, we offer, for completeness and without even attempting to be exhaustive, a representative synopsis of known experimental benchmarking results that highlight the ability of stability-based methods for model selection and the computational resources that they require for the task. As a whole, the contributions of this paper generalize in several respects reference methodologies in statistics and show that algorithmic approaches can yield deep methodological insights into this area, in addition to practical computational procedures. [Copyright &y& Elsevier]

Abstract: A mixture of Student t-distributions (MoT) has been widely used to model multivariate data sets with atypical observations, or outliers for robust clustering. In this paper, we developed a novel robust clustering approach by modeling the data sets using mixture of Pearson type VII distributions (MoP). An EM algorithm is developed for the maximum likelihood estimation of the model parameters. An outlier detection criterion is derived from the EM solution. Controlled experimental results on the synthetic datasets show that the MoP is more viable than the MoT. The MoP performs comparably if not better, on average, in terms of outlier detection accuracy and out-of-sample log-likelihood with the MoT. Furthermore, we compared the performances of the Pearson type VII and the student t mixtures on the classification of several real pattern recognition data sets. The comparison favours the developed Pearson type VII mixtures. [ABSTRACT FROM AUTHOR]

Abstract: This paper introduces a new linear regression method for interval valued-data. The method is based on the symmetrical linear regression methodology such that the prediction of the lower and upper bounds of the interval value of the dependent variable is not damaged by the presence of interval-valued data outliers. The method considers mid-points and ranges of the interval values assumed by the variables in the learning set. The prediction of the boundaries of an interval is accomplished through a combination of predictions from mid-point and range of the interval values. The evaluation of the method is based on the average behavior of a pooled root mean-square error. Experiments with real and simulated symbolic interval data sets demonstrate the usefulness of this symbolic symmetrical linear regression method. [ABSTRACT FROM AUTHOR]

Abstract: In an asymmetrical interval data set, the error ranges of the upper and lower interval ends are different. This situation is common in practice because of the usual presence of uncertain influences. In prior “crisp input and interval output” regression analysis, a crude symmetrical estimation is obtained, and the asymmetrical character of training data cannot be depicted exactly. In this paper, an asymmetrical interval data analysis is proposed for the first time. The two interval ends are studied independently, and a set of regression models and -insensitive functions are proposed to strengthen the description of the interval ends. The support vector machine (SVM) is imported into this approach (for its model-free character in nonlinear regression) and further extended by -insensitive functions to the extended -SVM. A robust algorithm is presented to eliminate the effect of outliers. Experiments are then presented to verify the quality of performance of the extended -SVM. Advantages over other approaches are considered in the conclusion. [Copyright &y& Elsevier]