Number of necessary training examples for Neural Networks with different number of trainable parameters.

In this work, the network complexity should be reduced with a concomitant reduction in the number of necessary training examples. The focus thus was on the dependence of proper evaluation metrics on the number of adjustable parameters of the considered deep neural network. The used data set encompassed Hematoxylin and Eosin (H&E) colored cell images provided by various clinics. We used a deep convolutional neural network to get the relation between a model's complexity, its concomitant set of parameters, and the size of the training sample necessary to achieve a certain classification accuracy. The complexity of the deep neural networks was reduced by pruning a certain amount of filters in the network. As expected, the unpruned neural network showed best performance. The network with the highest number of trainable parameter achieved, within the estimated standard error of the optimized cross-entropy loss, best results up to 30% pruning. Strongly pruned networks are highly viable and the classification accuracy declines quickly with decreasing number of training patterns. However, up to a pruning ratio of 40%, we found a comparable performance of pruned and unpruned deep convolutional neural networks (DCNN) and densely connected convolutional networks (DCCN).

Degradation of polyvinyl alcohol (PVA) by homogeneous and heterogeneous photocatalysis applied to the photochemically enhanced Fenton reaction.

The reaction mechanism of the oxidative degradation of polyvinyl alcohol (PVA) by the photochemically enhanced Fenton reaction was studied using a homogeneous (Fe2+(aq) + H2O2) and a heterogeneous reaction system (iron(III)-exchanged zeolite Y+ H2O2). In the homogeneous Fenton system, efficient degradation was observed in a batch reactor, equipped with a medium pressure mercury arc in a Pyrex envelope and employing 80% of the stoichiometric amount of H2O2 required for the total oxidation of PVA and a concentration ratio as low as I mole of iron(II) sulfate per 20 moles of PVA sub-units (C2H40). Model PVA polymers of three different molecular weights (15,000, 49,000 and 100,000 g mol(-1)) were found to follow identical degradation patterns. Strong experimental evidence supports the formation of supermacromolecules (MW: 1-5 x 10(6) g/mol) consisting of oxidized PVA and trapped iron(III) at an early reaction stage. Low molecular weight intermediates, such as oxalic acid, formic acid or formaldehyde were not found during PVA degradation in the homogeneous Fenton system, and we may deduce that the manifold of degradation reactions is mainly taking place within the super-macromolecules from which CO2 is directly released. However, in the heterogeneous Fenton system, the reaction behavior was found to be distinctly different: a decrease of the molecular weights of all three tested monodisperse PVA samples was observed by the broadening of the GPC-traces during irradiation, and oxalic acid was formed. The results lead to the mechanistic hypothesis that during the heterogeneous Fenton process, the cleavage of the PVA-chains may occur at random positions, the reactive centres being located inside the iron(III)-doped zeolite Y photocatalysts.

Optimal experimental design and artificial neural networks applied to the photochemically enhanced Fenton reaction.

Among advanced oxidation processes (AOPs), the photochemically enhanced Fenton reaction may be considered as one of the most efficient for the degradation of contaminants in industrial wastewater. This process involves a series of complex reactions. Therefore, an empirical model based on artificial neural networks has been developed for fitting the experimental data obtained in a laboratory batch reactor for the degradation of 2,4-dimethyl aniline (2,4-xylidine), chosen as a model pollutant. The model describes the evolution of the pollutant concentration during irradiation time as a function of the process conditions. It has been used for simulating the behavior of the reaction system in sensitivity studies aimed at optimizing the amounts of reactants employed in the process, an iron(III) salt and hydrogen peroxide, as well as the temperature. The results show that the process is most sensitive to the concentration of iron(III) salt and temperature, whereas the concentration of hydrogen peroxide has a minor effect.

An N,N'-dialkyl-4,4'-bipyridinium-modified titanium-dioxide photocatalyst for water remediation--observation and application of supramolecular effects in photocatalytic degradation of pi-donor organic compounds.

The photocatalytic activity of TiO2 (Degussa P-25) modified with a 4,4'-bipyridinium monolayer (V2+-TiO2) has been compared with that of conventional TiO2-P-25 by investigating the efficiency of degradation of a series of four organic model compounds with increasing pi-donor capacity (2,4-xylidine, 2,4-dimethylphenol, hydroquinone, and dimethylhydroquinone). As far as the mechanism of the first oxidation reaction is concerned, evidence for the formation of supramolecular donor-acceptor complexes with the bipyridinium units at the semiconductor surface was obtained by comparison of the Langmuir-adsorption characteristics and the efficiencies of photodegradation of the different substrates. Furthermore, the main intermediates of the photocatalytic degradation of 2,4-xylidine were identified, and the presence of 2,4-dimethylphenol indicates that the main pathway of substrate oxidation proceeds via electron transfer from the adsorbed organic substrate to the "holes" within the valence band of the photoexcited semiconductors V2+-TiO2 and TiO2. The efficiencies of photocatalytic degradation by both V2+-TiO2 and TiO2 were limited by the trapping efficiency of the conduction band electrons by molecular oxygen.