A dynamic over-sampling procedure based on sensitivity for multi-class problems

Classification with imbalanced datasets supposes a new challenge for researches in the framework of machine learning. This problem appears when the number of patterns that represents one of the classes of the dataset (usually the concept of interest) is much lower than in the remaining classes. Thus, the learning model must be adapted to this situation, which is very common in real applications. In this paper, a dynamic over-sampling procedure is proposed for improving the classification of imbalanced datasets with more than two classes. This procedure is incorporated into a memetic algorithm (MA) that optimizes radial basis functions neural networks (RBFNNs). To handle class imbalance, the training data are resampled in two stages. In the first stage, an over-sampling procedure is applied to the minority class to balance in part the size of the classes. Then, the MA is run and the data are over-sampled in different generations of the evolution, generating new patterns of the minimum sensitivity class (the class with the worst accuracy for the best RBFNN of the population). The methodology proposed is tested using 13 imbalanced benchmark classification datasets from well-known machine learning problems and one complex problem of microbial growth. It is compared to other neural network methods specifically designed for handling imbalanced data. These methods include different over-sampling procedures in the preprocessing stage, a threshold-moving method where the output threshold is moved toward inexpensive classes and ensembles approaches combining the models obtained with these techniques. The results show that our proposal is able to improve the sensitivity in the generalization set and obtains both a high accuracy level and a good classification level for each class.     

A comment on: “Fast and numerically stable methods for the computation of Zernike moments” by Singh et al. [Pattern Recognition, 43(2010), Pages 2497-2506]

The aim of the present comment is to point out that the q-Recursive relation introduced by Singh et al. is wrong. These errors have been corrected in the comment. The numerical experiments show that the correct recurrence relation is consistent with the original q-Recursive method.     

Minimizing a linear objective function under a fuzzy max-t norm relation equation constraint

The work examines the feasibility of minimizing a linear objective function subject to a max-t fuzzy relation equation constraint, where t is a continuous/Archimedean t-norm. Conventional methods for solving this problem are significantly improved by, first separating the problem into two sub-problems according to the availability of positive coefficients. This decomposition is thus more easily handled than in previous literature. Next, based on use of the maximum solution of the constraint equation, the sub-problem with non-positive coefficients is solved and the size of the sub-problem with positive coefficients reduced as well. This step is unique among conventional methods, owing to its ability to determine as many optimal variables as possible. Additionally, several rules are developed for simplifying the remaining problem. Finally, those undecided optimal variables are obtained using the covering problem rather than the branch-and-bound methods. Three illustrative examples demonstrate that the proposed approach outperforms conventional schemes. Its potential applications are discussed as well.     

Computational performance of a parallelized three-dimensional high-order spectral element toolbox

In this paper, a comprehensive performance review of an MPI-based high-order three-dimensional spectral element method C++ toolbox is presented. The focus is put on the performance evaluation of several aspects with a particular emphasis on the parallel efficiency. The performance evaluation is analyzed with the help of a time prediction model based on a parameterization of the application and the hardware resources. Two tailor-made benchmark cases in computational fluid dynamics (CFD) are introduced and used to carry out this review, stressing the particular interest for clusters with up to thousands of cores. Some problems in the parallel implementation have been detected and corrected. The theoretical complexities with respect to the number of elements, to the polynomial degree, and to communication needs are correctly reproduced. It is concluded that this type of code has a nearly perfect speedup on machines with thousands of cores, and is ready to make the step to next-generation petaFLOP machines.     

Laminar convection in the annulus of a double-pipe with triangular fins

A steady and laminar convective flow has been numerically simulated in the fully-developed annular region of a finned double-pipe subjected to the constant heat flux boundary condition imposed at the inner-pipe wall. Finite element method has been employed in this study. Friction factor and Nusselt number have been studied as flow characteristics against variations in the ratio of radii of the inner and the outer pipes, the fin height, the fin half angle and the number of fins. The results show significant enhancement in the heat transfer rate in both the cases when sufficient pumping power is available and when it is not. The minimum and maximum increase in the product of friction factor and Reynolds number relative to the finless geometry is more than one time and more than 40 times respectively while gain in the relative value of Nusselt number lies in the range 1–177. This provides an evidence of more than four times enhancement in the heat transfer coefficient relative to that in the pressure loss as a result of extended fin surfaces.     

A new solution for a queueing model of a manufacturing cell with negative customers under a rotation rule

In this paper, we consider a queueing model extension for a manufacturing cell composed of a machining center and several parallel downstream production stations under a rotation rule. A queueing model is extended with the arrival processes of negative customers to capture failures of production stations, reorganization of works and disasters in the manufacturing cell. We present an exact solution for the steady-state probabilities of the proposed queueing model. The solution does not require the approximation of the infinite sum. In addition, it provides an alternative way to compute the rate matrix for the matrix-geometric method as well.     

Training linear ranking SVMs in linearithmic time using red-black trees

We introduce an efficient method for training the linear ranking support vector machine. The method combines cutting plane optimization with red-black tree based approach to subgradient calculations, and has O(ms + mlog (m)) time complexity, where m is the number of training examples, and s the average number of non-zero features per example. Best previously known training algorithms achieve the same efficiency only for restricted special cases, whereas the proposed approach allows any real valued utility scores in the training data. Experiments demonstrate the superior scalability of the proposed approach, when compared to the fastest existing RankSVM implementations.     

Bayesian analysis of the patterns of biological susceptibility via reversible jump MCMC sampling

In some biological experiments, it is quite common that laboratory subjects differ in their patterns of susceptibility to a treatment. Finite mixture models are useful in those situations. In this paper we model the number of components and the component parameters jointly, and base inference about these quantities on their posterior probabilities, making use of the reversible jump Markov chain Monte Carlo methods. In particular, we apply the methodology to the analysis of univariate normal mixtures with multidimensional parameters, using a hierarchical prior model that allows weak priors while avoiding improper priors in the mixture context. The practical significance of the proposed method is illustrated with a dose-response data set.     

Modelling acute renal failure using blood and breath biomarkers in rats

This paper compares three methods for estimating renal function, as tested in rats. Acute renal failure (ARF) was induced via a 60-min bilateral renal artery clamp in 8 Sprague-Dawley rats and renal function was monitored for 1 week post-surgery. A two-compartment model was developed for estimating glomerular filtration via a bolus injection of a radio-labelled inulin tracer, and was compared with an estimated creatinine clearance method, modified using the Cockcroft-Gault equation for rats. These two methods were compared with selected ion flow tube-mass spectrometry (SIFT-MS) monitoring of breath analytes. Determination of renal function via SIFT-MS is desirable since results are available non-invasively and in real time. Relative decreases in renal function show very good correlation between all 3 methods (R² = 0.84, 0.91 and 0.72 for breath-inulin, inulin-creatinine, and breath-creatinine correlations, respectively), and indicate good promise for fast, non-invasive determination of renal function via breath testing.     

紧密衔接工序组联动的综合调度算法

由于现有的工序间存在紧密衔接条件的复杂产品综合调度问题, 采用的移动交换算法不易于软件实现且没有考虑移动工序后产生的连锁反应引起较高算法复杂度的问题, 提出将具有紧密衔接约束条件的工序组进行统一联动的综合调度算法.该算法利用将具有紧密衔接约束条件的工序分组的扩展加工工艺树模型, 按路径上属于工序组的工序个数多少确定所在路径工序组调度的次序, 通过降低对工序组的限制要求降低算法复杂度; 对于被调度工序组中各工序的前序工序, 按工序组中工序的加工顺序确定调度次序, 对某个工序的前序工序采用复杂度较低的拟关键路径法确定工序的调度次序; 调度完所有紧密衔接工序组后, 剩余的标准工序按拟关键路径法确定调度顺序; 采取工序首次适应调度算法调度标准工序和工序组, 由于工序组中工序采取按序紧密衔接的联动调度方式确定工序组的开始时间, 避免了二次调整, 进一步降低了算法复杂度. 分析和实例表明, 所提出的综合算法比以往算法复杂度更低, 调度结果更优且更易于实现.