Multiobjective study to implement outpatient appointment system at Hedi Chaker Hospital

The outpatient department at Hedi Chaker Hospital in Sfax, Tunisia, is currently using a block scheduling appointment whereby all patients are required to come at the beginning of the session. This brings about a tremendous waste of their time, creates congestion and puts further pressure on the medical staff. This paper evaluates a number of outpatient scheduling schemes based on the most recent research findings. Each model uses an appointment system scheme for scheduling with a well-performing appointment rule crossed with a sequencing rule. Environmental factors, such as the probability of no-shows and the probability of walk-ins, are included in the simulation of each of the different scenarios. Key measures of performance are determined for each of the different scenarios: waiting time in queues for each class of patients and resource utilisation. A Goal Programming Model is developed to choose the most appropriate schedule that optimizes the measures. Management preferences are incorporated to obtain a solution in line with their satisfaction.     

Context-dependent kernels for object classification

Kernels are functions designed in order to capture resemblance between data and they are used in a wide range of machine learning techniques, including support vector machines (SVMs). In their standard version, commonly used kernels such as the Gaussian one show reasonably good performance in many classification and recognition tasks in computer vision, bioinformatics, and text processing. In the particular task of object recognition, the main deficiency of standard kernels such as the convolution one resides in the lack in capturing the right geometric structure of objects while also being invariant. We focus in this paper on object recognition using a new type of kernel referred to as "context dependent.” Objects, seen as constellations of interest points, are matched by minimizing an energy function mixing 1) a fidelity term which measures the quality of feature matching, 2) a neighborhood criterion which captures the object geometry, and 3) a regularization term. We will show that the fixed point of this energy is a context-dependent kernel which is also positive definite. Experiments conducted on object recognition show that when plugging our kernel into SVMs, we clearly outperform SVMs with context-free kernels.     

Wind load modeling for topology optimization of continuum structures

Topology optimization of two and three dimensional structures subject to dead and wind loading is considered. The wind loading is introduced into the formulation by using standard expressions for the drag force, and a strategy is devised so that wind pressure is ignored where there is no surface obstructing the wind. A minimum compliance design formulation is constructed subject to a volume constraint using the Solid Isotropic Material with Penalization model. The optimization problem is solved using the Method of Moving Asymptotes, modified by including a line search and by changing the formula for the update of asymptotes. To obtain black/white design, intermediate density values, which are used as design variables, are controlled by imposing an explicit constraint. Numerical examples of a windmill structure demonstrate that the proposed formulation rationally incorporates the effect of wind loading into the topology optimization problem as illustrated by void appearing in the optimal structure.     

Contributing vertices-based Minkowski sum computation of convex polyhedra

Minkowski sum is an important operation. It is used in many domains such as: computer-aided design, robotics, spatial planning, mathematical morphology, and image processing. We propose a novel algorithm, named the Contributing Vertices-based Minkowski Sum (CVMS) algorithm for the computation of the Minkowski sum of convex polyhedra. The CVMS algorithm allows to easily obtain all the facets of the Minkowski sum polyhedron only by examining the contributing vertices—a concept we introduce in this work, for each input facet. We exploit the concept of contributing vertices to propose the Enhanced and Simplified Slope Diagram-based Minkowski Sum (ESSDMS) algorithm, a slope diagram-based Minkowski sum algorithm sharing some common points with the approach proposed by Wu et al. [Wu Y, Shah J, Davidson J. Improvements to algorithms for computing the Minkowski sum of 3-polytopes. Comput Aided Des. 2003; 35(13): 1181-92]. The ESSDMS algorithm does not embed input polyhedra on the unit sphere and does not need to perform stereographic projections. Moreover, the use of contributing vertices brings up more simplifications and improves the overall performance. The implementations for the mentioned algorithms are straightforward, use exact number types, produce exact results, and are based on CGAL, the Computational Geometry Algorithms Library. More examples and results of the CVMS algorithm for several convex polyhedra can be found at http://liris.cnrs.fr/hichem.barki/mksum/CVMS-convex.     

A particular Gaussian mixture model for clustering and its application to image retrieval

We introduce a new method for data clustering based on a particular Gaussian mixture model (GMM). Each cluster of data, modeled as a GMM into an input space, is interpreted as a hyperplane in a high dimensional mapping space where the underlying coefficients are found by solving a quadratic programming (QP) problem. The main contributions of this work are (1) an original probabilistic framework for GMM estimation based on QP which only requires finding the mixture parameters, (2) this QP is interpreted as the minimization of the pairwise correlations between cluster hyperplanes in a high dimensional space and (3) it is solved easily using a new decomposition algorithm involving trivial linear programming sub-problems. The validity of the method is demonstrated for clustering 2D toy examples as well as image databases.     

Lithium recovery from highly concentrated solutions: Response surface methodology (RSM) process parameters optimization

Lithium recovery from aqueous solutions was studied by trapping lithium ions using a gel of aluminum hydroxide freshly prepared in situ under the reaction of a strong base. A 2^4 − 1 fractional factorial design and central composite design were employed for experimental design and analysis of the results. The combined effect of molar ratio Al/Li, pH and stirring time on lithium trapping at 25 °C was investigated and optimized using response surface methodology (RSM). The optimum values of these factors were found to be 4.7, 7.2 and 3 h respectively. In this case the lithium trapping percent is ∼ 95%. The obtained model is highly significant (F_obs > F_tabulate and low p-value) with a correlation coefficient of 96.64%. On the other hand, linear, quadratic and interaction terms in this model have the largest statistical effect on the response (confidence level = 99.9%).     

基于动态贝叶斯网络的大词汇量连续语音识别和音素切分研究

提出一个新颖的单流多状态动态贝叶斯网络(Single stream Multi-states Dynamic Bayesian Network, SM-DBN)模型,以实现大词汇量连续语音识别和音素切分.该模型在Bil m es等人提出的单流动态贝叶斯网络(Single stream Dynamic Bayesian Network, Phone-sh ared,SS-DBN-P)模型(识别基元为词)基础上,增加了一个隐含的状态节点层,每个词由它的对 应音素组成,而音素采用固定个数的状态描述,状态和观测向量直接连接.它的识别基元为音 素,描述了音素的动态发音变化过程.大词汇量语音识别的实验结果表明:在纯净语音环境下 ,SM-DBN模型的识别率比HMM和SS-DBN-P模型的识别率分别提高了13.01%和35.2%,而 音频流的音素切分正确率则分别提高了10%和44%.     

A hybrid optical–mechanical calibration procedure for the Scalable-SPIDAR haptic device

In this research, a simple, yet, efficient calibration procedure is presented in order to improve the accuracy of the Scalable-SPIDAR haptic device. The two-stage procedure aims to reduce discrepancies between measured and actual values. First, we propose a new semi-automatic procedure for the initialization of the haptic device. To perform this initialization with a high level of accuracy, an infrared optical tracking device was used. Furthermore, audio and haptic cues were used to guide the user during the initialization process. Second, we developed two calibration methods based on regression techniques that effectively compensate for the errors in tracked position. Both neural networks and support vector regression methods were applied to calibrate the position errors present in the haptic device readings. A comparison between these two regression methods was carried out to show the underlying algorithm and to indicate the inherent advantages and limitations for each method. Initial evaluation of the proposed procedure indicated that it is possible to improve accuracy by reducing the Scalable-SPIDAR’s average absolute position error to about 6 mm within a 1 m × 1 m × 1 m workspace.     

A reinforcement learning approach for UAV target searching and tracking

Multimedia Tools and Applications - Owing to the advantages of Unmanned Aerial Vehicle (UAV), such as the extendibility, maneuverability and stability, multiple UAVs are having more and more...