Local feature selection for multiple instance learning

Journal of Intelligent Information Systems - We propose a local feature selection method for the Multiple Instance Learning (MIL) framework. Unlike conventional feature selection algorithms that...     

A hybrid approach based on SERVQUAL and fuzzy TOPSIS for evaluating transportation service quality

Managing service quality is vital to retain customer satisfaction and augment revenues for any business organization. Often it is difficult to assess service quality due to lack of quantifiable measures and limited data. In this paper, we present a hybrid approach based on SERVQUAL and fuzzy TOPSIS for evaluating service quality of urban transportation systems. The proposed approach consists of three steps. The first step involves development of a SERVQUAL based questionnaire to collect data for measuring transportation service quality. The participants provide linguistic assessments to rate the service quality criteria and the alternatives. In step 2, the linguistic ratings are combined through fuzzy TOPSIS to generate an overall performance score for each alternative. The alternative with the highest score is finally chosen. In step 3, sensitivity analysis is conducted to evaluate the influence of criteria weights on the decision making process.The strength of the proposed approach is its practical applicability and ability to provide solution under partial or lack of quantitative information. An application of the proposed approach for evaluation of service quality of metro in Montreal is provided.     

A wideband rectifier array on dual‐polarized differential‐feed fractal slotted ground antenna for RF energy harvesting

This article reports a novel wideband rectenna for RF energy harvesting applications. A wideband fractal slotted ground antenna (SGA) is adopted. The operating frequency bands of the antenna are GSM, UMTS, Wi‐Fi, and LTE2600/4G bands. The antenna is fed by a dual‐polarized and differential‐feed (DP‐DF) microstrip lines disposed with an angle of 90° each relative to the other. The feed lines are etched on the bottom side of the substrate and connected to an array of four wideband RF‐to‐DC rectifiers. A nonuniform transmission lines filter ensures wideband behavior for each rectifier. The rectenna performances are simulated and measured. The experiments show an output DC voltage of 1 V at a power density of 26.6 μW/cm² over the frequency band of operation with a peak efficiency of 50%. The proposed rectenna is suitable for energy harvesting applications in urban environments.     

基于动态贝叶斯网络的语音识别及音素切分研究

研究了一种基于动态贝叶斯网络(dynamic bayesian networks, DBN)的语音识别建模方法,利用GMTK(graphical model tool kits)工具构建音素级音频流DBN语音训练和识别模型,同时与传统的基于隐马尔可夫的语音识别结果进行比较,并给出词与音素的切分结果.实验表明,在各种信噪比测试条件下,基于DBN的语音识别结果与基于HMM的语音识别结果相当,并表现出一定的抗噪性,音素的切分结果也比较准确.     

Parameter identification and computational simulation of polyurethane foaming process by finite pointset method

Numerical simulation of the polyurethane foaming process is a valuable method to analyze the molding process at an early stage of product development to shorten time-to-market cycles and cut costs by using fewer prototypes. However, this process involves highly coupled thermo-chemo-rheological modeling and needs adequate model parameters’ identification. A theoretical model including chemical reactions and thermo-rheological coupling of conservation equations was developed. Based on the theoretical model, three-dimensional numerical simulation for mold filling of the polyurethane foam was carried out by using Finite Pointset Method (FPM) to predict flow field, flow front advancement, temperature and density distributions during mold filling. A FOAMAT system was used to monitor foam height rise and reaction temperature on a cylindrical test tube and foam viscosity was measured by using a dynamic rotational rheometer with parallel-plate system. The parameters of the model were identified by an inverse analysis method which consists in determining the parameters by comparing the computed quantities to those measured experimentally. The overall modeling was validated by using short shot foams obtained with a panel mold cavity. Mold filling of an automotive underlay carpet cavity was investigated numerically. Flow front results were successfully compared to short shot foams obtained with the industrial mold cavity.     

Experimental study of the performance of a cooling tower used in a solar distiller

In this work, we investigated experimentally the thermal performance of a forced cooling tower used in a solar desalination system based on humidification-dehumidification of air. The cooling tower is a counter flow wet one filled with film packing materials.The measured variables were obtained for wide ranges of mass flow rates of air and water as well as for several inlet water temperatures; the tower characteristic and efficiency were then evaluated and expressed in terms of water to air mass flow rate ratio.     

Bayesian Unsupervised Learning for Source Separation with Mixture of Gaussians Prior

This paper considers the problem of source separation in the case of noisy instantaneous mixtures. In a previous work [1], sources have been modeled by a mixture of Gaussians leading to an hierarchical Bayesian model by considering the labels of the mixture as i.i.d hidden variables. We extend this modelization to incorporate a Markovian structure for the labels. This extension is important for practical applications which are abundant: unsupervised classification and segmentation, pattern recognition and speech signal processing. In order to estimate the mixing matrix and the a priori model parameters, we consider observations as incomplete data. The missing data are sources and labels: sources are missing data for observations and labels are missing data for incomplete missing sources. This hierarchical modelization leads to specific restoration maximization type algorithms. Restoration step can be held in three different manners: (i) Complete likelihood is estimated by its conditional expectation. This leads to the EM (expectation-maximization) algorithm [2], (ii) Missing data are estimated by their maximum a posteriori. This leads to JMAP (Joint maximum a posteriori) algorithm [3], (iii) Missing data are sampled from their a posteriori distributions. This leads to the SEM (stochastic EM) algorithm [4]. A Gibbs sampling scheme is implemented to generate missing data. We have also introduced a relaxation strategy into these algorithms to reduce the computational cost which is due to the exponential influence of the number of source components and the number of the mixture Gaussian components.