A sample-level DCNN for music auto-tagging

Multimedia Tools and Applications - Deep convolutional neural networks (DCNNs) has been widely used in music auto-tagging which is a multi-label classification task that predicts tags of audio...     

Fabrication and characterization of polyvinyl chloride mixed matrix membranes containing high aspect ratio anatase titania and hydrous manganese oxide nanoparticle for efficient removal of heavy metal ions: Competitive removal study

In this study, novel polyvinyl chloride (PVC) ultrafiltration mixed matrix membranes (MMMs) containing high aspect ratio anatase titania (ANT) and hydrous manganese oxide (HMO) nanoparticles were synthesized in order to decontaminate cadmium and copper metal ions. ANT and HMO nanoparticles with various loadings in a range of 5-15 were used in the casting solution of MMMs. The characterization of fabricated MMMs was carried out with respect to the structural morphology, hydrophilicity, and composition by scanning electron microscopy (SEM), water contact angle, and Fourier transform infrared spectroscopy (FTIR), respectively. With an increase in the nanoparticle loadings, the change in other membrane characteristics such as mean pore size, pure water flux, and porosity were also evaluated. The results revealed that the mean pore size and water flux increased at a higher loading of ANT and HMO nanoparticles, while the contact angle and porosity of the membranes showed reverse trends. Moreover, the higher flux of pure water was obtained for PVC-ANT MMMs compared to PVC-HMO MMMs because of the larger mean pore size, higher porosity, and hydrophilicity of PVC-ANT MMMs. In this study, the decontamination of cadmium and copper metal ions in single and binary systems of heavy metals was investigated and the effect of Mn²⁺ (as an interfering ion) was also studied. The evaluation of the metal ion removal data demonstrated that the affinity sequence of both the PVC-ANT-15 and PVC-HMO-15 MMMs for heavy metal removal was Cu²⁺ > Cd²⁺ > Mn²⁺ in the single and binary systems. It was found through ultrafiltration (UF) experiments that PVC-ANT MMM was the most efficient membrane in the elimination of heavy metals due to the superior ANT adsorption capacity. However, the overall findings disclosed that both ANT and HMO nanoparticles are suitable candidates for the preparation of MMMs used in cadmium and copper decontamination from aqueous solutions.     

Modeling and solving a multi-objective supplier quota allocation problem considering transaction costs

Suppliers’ evaluation is a subject, which has attracted the attention of many researchers. The performance of potential suppliers is evaluated against multiple criteria rather than considering a single factor such as cost or quality. One of the major objectives of suppliers’ evaluation is to determine the optimal quota assigned to each supplier while needing to replenish an order. This problem has been studied by many researchers as a multi-objective problem. The usual objectives are minimizing the purchasing cost, rejected units, and late delivered units. However, in a few researches maximizing the evaluation scores of the selected suppliers is considered as fourth objective. In this paper, we present a model with five objectives including minimizing the transaction costs of purchasing from suppliers as well as the four addressed objectives. We convert the model to a single objective one using the well-known weighting method, solve it utilizing two meta-heuristic algorithms, and analyze the efficiency of the heuristics. The reason why we utilize the meta-heuristic algorithms is that the problem is proved to be an NP-hard one.     

Effects of flow history on oil entrapment in porous media: An experimental study

The effect of flow history on fluid phase entrapment during immiscible two‐phase flow in Hele‐Shaw cells packed with spherical and crushed glass beads is investigated. The wetting fluid is injected into an initially oil saturated cell at a well‐defined capillary number. It is observed that the size and shape of the trapped clusters strongly depend on the history of flooding such that less oil was trapped in the medium when the injecting capillary number gradually increased to the final maximum capillary number compared to the case when the injection was started and maintained constant at the maximum capillary number. In addition, a comprehensive series of experiments were conducted to delineate the effects of the capillary number on the phase entrapment. Contrary to previously published data, our experimental data reveal that the residual oil saturation depends on capillary number nonmonotonically. A physically based relationship to scale the capillary desaturation curve is proposed. © 2014 American Institute of Chemical Engineers AIChE J, 61: 1385–1390, 2015     

A statistical approach to engineer a biocomposite formulation from biofuel coproduct with balanced properties

A 3²‐full factorial design of experiment (DOE) and regression modeling were implemented together as a practical approach to attain a distillers' grains‐filled biocomposite with balanced mechanical and physical properties. The effects of compatibilizer and lubricant on tensile strength, flexural modulus, impact strength and melt flow index of the biocomposites were studied. Analysis of variance (ANOVA) was implemented to develop least square regression models containing statistically significant main effects (linear and quadratic) and interaction effect. The developed models showed good predictability for the new measurements. The statistical approach adopted in this work including overlaying contour plots of the response surfaces in the studied level domain was effective in highlighting an optimized region that leads to balanced mechanical and physical properties. © 2014 The Authors Journal of Applied Polymer Science Published by Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2014 , 131, 40443.     

NBTI and hot-carrier effects in accumulation-mode Pi-gate pMOSFETs

Negative bias temperature instability (NBTI) and hot-carrier induced device degradation in accumulation-mode Pi-gate pMOSFETs have been studied for different fin widths ranging from 20 to 40 nm. The NBTI induced device degradation is more significant in narrow devices. This result can be explained by enhanced diffusion of hydrogen at the corners in multiple-gate devices. Due to larger impact ionization, hot-carrier induced device degradation is more significant in wider devices. Finally, hot-carrier induced device degradation rate is highest under stress conditions where V_GS = V_TH.     

Comparison of different surface orientation in narrow fin MuGFETs

Device performance characteristics are investigated for different surface orientation and doping concentration on accumulation-mode p-type and inversion-mode n-type MuGFETs. Short-channel effects and drain breakdown voltage are better is carrier transport is in the (1 0 0) direction than in the (1 1 0) direction. This is due to the larger Si/SiO₂ interface roughness, the higher density of interface state at (1 1 0) surfaces, and to the difference of effective mass. The mobility in PMOS devices, however, is much higher in the (1 1 0) direction than that in the (1 0 0) direction. For better performance of device, our results show that optimized fin orientation can improve device stability and performance.     

Numerical simulation of planar shear flow passing a rotating cylinder at low Reynolds numbers

Two-dimensional shear flow over a rotating circular cylinder is investigated using lattice Boltzmann method. Simulations are performed at a fixed blockage ratio (B = 0.1) while the Reynolds number, nondimensional shear rate (K) and absolute rotational speed range as 80 ≤ Re ≤ 180, 0 ≤ K ≤ 0.2 and −2 ≤ β ≤ 2, respectively. To verify the simulation, the results are compared to previous experimental and numerical data. Quantitative information about the flow variables such as drag and lift coefficients, pressure coefficient and vorticity distributions on the cylinders is highlighted. It is found that, generally, with the increment in |β|, the absolute value of time average lift coefficient increases and time average drag coefficient decreases, and beyond a certain magnitude of β, the vortex shedding vanishes. It is also revealed that the drag coefficient decreases as the Reynolds number increases while the effect of the Reynolds number on lift is almost negligible. At the end, correlations for drag and lift coefficients ([`(C<sub>D</sub> )] ,[`(C_L )]){(\overline {C_D } ,\overline {C_L })} are extracted from the numerical data.     

Terahertz graphene optics

The magnitude of the optical sheet conductance of single-layer graphene is universal, and equal to e ²/4? (where 2??? = h (the Planck constant)). As the optical frequency decreases, the conductivity decreases. However, at some frequency in the THz range, the conductivity increases again, eventually reaching the DC value, where the magnitude of the DC sheet conductance generally displays a sample- and doping-dependent value between ??e ²/h and 100 e ²/h. Thus, the THz range is predicted to be a non-trivial region of the spectrum for electron transport in graphene, and may have interesting technological applications. In this paper, we present the first frequency domain measurements of the absolute value of multilayer graphene (MLG) and single-layer graphene (SLG) sheet conductivity and transparency from DC to 1 THz, and establish a firm foundation for future THz applications of graphene.      

Boost-wise pre-loaded mixture of experts for classification tasks

A modified version of Boosted Mixture of Experts (BME) is presented in this paper. While previous related works, namely BME, attempt to improve the performance by incorporating complementary features of a hybrid combining framework, they have some drawback. Analyzing the problems of previous approaches has suggested several modifications that have led us to propose a new method called Boost-wise Pre-loaded Mixture of Experts (BPME). We present a modification in pre-loading (initialization) procedure of ME, which addresses previous problems and overcomes them by employing a two-stage pre-loading procedure. In this approach, both the error and confidence measures are used as the difficulty criteria in boost-wise partitioning of problem space.