Feature Fusion Models for Deep Autoencoders: Application to Traffic Flow Prediction

ABSTRACT

Due to reduction in dimensionality and extraction of the definitive features of input data, deep architectures have achieved significant success in various machine learning applications. Considering their successful applications in speech recognition and image classification, the main goal of this research is to investigate the performance of the sparse autoencoders utilized in regression analysis. To this end, deep sparse autoencoders with the standard method of training, cascaded, and partially cascaded architectures, fed with the fusion of low- and high-level features, are proposed and implemented. The regression task is to forecast the vehicular flow rate of a location on an arterial highway using different traffic variables of several locations ahead in the Twin Cities Metro area of Minneapolis. The results demonstrate that the partially cascaded model exhibits advancements in yielding more accurate results than the other two architectures fed with the features that correlate the most to the traffic flow rate.     

Aqueous free-radical polymerization of PEGMEMA macromer: kinetic studies via an on-line 1H NMR technique

Free-radical polymerization of polyethylene glycol methyl ether methacrylate macromer (PEGMEMA) was studied in aqueous media and in the presence of potassium persulfate (KPS) as water soluble initiator. An on-line nuclear magnetic resonance (NMR) method was applied to record the reaction data and determine the monomer conversion at various times during the polymerization reaction progress. ¹H NMR spectrum of reaction mixture containing monomer, initiator and resultant polymer was continuously recorded in NMR instrument with the increase of reaction time. By processing the obtained data from NMR spectrum, the rate equation can be derived and reaction order can be determined with regard to monomer and initiator concentration. In other words, to determine the order of polymerization with regard to the concentration of reactants in free-radical polymerization of PEGMEMA, macromer samples with different amounts of monomer and KPS were prepared and polymerized at 50?°C. Orders of reaction with respect to monomer and initiator molar concentrations were equal to 1.025 and 0.480, respectively. The obtained values for reaction orders in this study were consistent with the classical kinetic rate equation in which the dependency of polymerization rate (R _p) on monomer and initiator concentrations was equal to 1 and 0.5, respectively. To measure polymerization activation energy (E _a), the effect of reaction temperature on the polymerization rate was investigated and E _a?=?37.08?kJ/mol was obtained at the temperature range of 40?C50?°C.     

Understanding the molecular mechanism of thio-Claisen rearrangement of allyl phenyl sulfide and allyl vinyl sulfide using bonding evolution theory coupled with NCI analysis

A joint application of bonding evolution theory (BET) and non-covalent interaction (NCI) analysis provides a powerful tool to explain the molecular mechanism of organic reactions. Thio-Claisen rearrangements (TCR) of allyl phenyl sulfide and allyl vinyl sulfide have been studied energetically using the MPWB1K/6-311G(2d,d,p), M06-2x/6-311G(2d,d,p) and the complete basis set model chemistry CBS-QB3 to predict activation barriers and thermodynamic parameters. Loss of aromaticity within the benzene ring along the TCR of allyl phenyl sulfide leads to a higher activation barrier and its endergonicity (endothermicity) nature, while TCR of allyl vinyl sulfide is an exergonic (exothermic) rearrangement with lower activation barrier. The molecular mechanisms are described by using MPWB1K/6-311G(2d,d,p) wavefunctions and energy profiles. The TCR of allyl phenyl sulfide and allyl vinyl sulfide can be described by the sequence of catastrophes 6-[C^?C]F[F^?F^?F]^TS[C^?C]C^?-0 and 8-C^?CC[F^?F^?F]^TSFCC^?-0, respectively, by the following chemical events: (a) homolytic bond breaking of C?S single bond and formation of pseudoradical centers on the atoms; (b) annihilation of the pseudoradical centers and reorganization in covalent bonds; (c) appearance of the pseudoradical centers on the terminal carbon atoms; (d) formation of new C?C single bond by the C-to-C coupling of pseudoradical centers when the TS of the reaction is reached and left behind. NCI isosurface appears in the region of broken C?S bond and forming C?C bond where the monosynaptic basins were localized.     

A magnetic adsorbent based on salicylic acid-immobilized magnetite nano-particles for pre-concentration of Cd(II) ions

In this research,an eco-friendly magnetic adsorbent based on Fe₃O₄/salicylic acid nanocomposite was fabricated using a facile one-pot co-precipitation method.The crystalline and morphological characterization of the prepared nanocomposite was performed by field emission scanning electron microscopy,X-ray diffraction,and Fourier transform infrared spectroscopy.The nanocomposite was employed as a magnetic solid-phase extraction agent for separation of Cd(II)ions from synthetic solutions.Some experimental factors affecting the extraction efficiency were investigated and optimized.Following elution with acetic acid(pH 3.5),the pre-concentrated analyte was quantified by flame atomic absorption spectrometry.In optimal conditions,a linear calibration graph was achieved in the concentration range of 0.2–30 ng·mL^?1 with a determination coefficient(R²)of 0.9953.The detection limit,the enhancement factor,inter-and intra-day relative standard deviations(for six consecutive extractions at the concentration level of 10 ng·mL^?1)were 0.04 ng·mL^?1,100,2.38%and 1.52%,respectively.To evaluate the accuracy of the method,a certified reference material(NIST SRM 1643e)was analyzed,and there was a good agreement between the certified and the measured values.It was successfully utilized to determine cadmium in industrial wastewater samples and the attained relative recovery values were between 96.8%and 103.2%.