Sharif Digital Repository

The approximation power of general feedforward neural networks with piecewise linear activation functions is investigated. First, lower bounds on the size of a network are established in terms of the approximation error and network depth and width. These bounds improve upon state- of-the-art bounds for certain classes of functions, such as strongly convex functions. Second, an upper bound is established on the difference of two neural networks with identical weights but different activation functions. © The Author(s) 2018  

In this paper we propose a new method for image noise reduction based on wavelet transform. In this method we: introduce an improved version of thresholding neural networks. (TNN) by utilizing a new class of smooth nonlinear thresholding functions as the activation function. Using this approach we will find the best thresholds in the sense of minimum mean square error (MMSE). Then using, TNN with obtained thresholds, we employ a cycle-spinningbased technique to reduce image artifacts. Experimental results indicate that the proposed method outperforms several other established wavelet denoising techniques, in terms of Peak-Signal-to-Noise-Ratio (PSNR) and visual quality. © 2007 IEEE  

Chitosan was crosslinked with different amount of glutaraldehyde to prepare appropriate hydrogels to be used as drug delivery system. The swelling behavior of freeze-dried hydrogels in aqueous media at different temperature and pHs has been examined. The swelling, porosity and biocompatibility behavior of samples were investigated to check effects of polymer/polymer and polymer/drug interactions on these system characteristics. Obtained experimental results illustrates that with increasing crosslinking agent from 0.068 to 0.30, swelling of the prepared samples degrees from 1200% to 600% and pore diameters change from 100 to 500 μm. To investigate systems biocompatibility in gastric... 

In this work, the silver particle absorption and antibacterial activity of cotton fabric when modified by low temperature plasma were investigated. The modification consisted of plasma pre-functionalization followed by one-step wet treatment with silver nitrate solution. Oxygen and nitrogen were used as the working gases in the system, and the results were compared. The results showed that nitrogen plasma-treated samples can absorb more silver particles than oxygen-treated samples, and thus the antibacterial activity of the samples in this case, which was analyzed by the counting bacteria test, was increased considerably  

A titanium dioxide nanotube (TNT) and CuO-doped TNT (TNT-CuO) were synthesized using a hydrothermal method and incorporated into carrageenan-based films. The SEM results confirmed the formation of uniform nanocomposite films. The addition of nanoparticles imparted UV-blocking properties to the carrageenan film and increased the mechanical strength, surface hydrophobicity, and water vapor barrier properties. The modified TiO2 (TNT and TNT-CuO)-incorporated carrageenan films showed significantly higher antibacterial activity than the TiO2-added film under visible light. Bananas packaged with the neat carrageenan and TiO2-added films were degraded considerably after 12 days of storage at 20 °C.... 

Sensors are one of the most vital instruments in Nuclear Power Plants (NPPs), and operators and safety systems monitor and analyze various parameters reported by them. Failure to detect sensors malfunctions or anomalies would lead to the considerable consequences. In this research, a new method based on thermal–hydraulic simulation by RELAP5 code and Feed-Forward Neural Networks (FFNN) is introduced to detect faulty sensors and estimate their correct value. For design an efficient neural net, seven feature selectors (i.e., Information gain, ReliefF, F-regression, mRMR, Plus-L Minus-R, GA, and PSO), three sigmoid activation functions (i.e., Logistic, Tanh and Elliot), and three training... 

Optical Neural Networks (ONNs) can be promising alternatives for conventional electrical neural networks as they offer ultra-fast data processing with low energy consumption. However, lack of suitable nonlinearity is standing in their road of achieving this goal. While this problem can be circumvented in feed-forward neural networks, the performance of the recurrent neural networks (RNNs) depends heavily on their nonlinearity. In this paper, we first propose and numerically demonstrate a novel reconfigurable optical activation function, named ROA, based on adding or subtracting the outputs of two saturable absorbers (SAs). RAO can provide both bounded and unbounded outputs by facilitating an... 

Citric acid was polymerized onto the surface of functionalized multiwall carbon nanotubes (MWCNTCOOH) and MWCNT-graft-poly(citric acid) (MWCNTg-PCA) hybrid materials were obtained. Due to the grafted poly(citric acid) branches, MWCNT-g-PCA hybrid materials not only were soluble in water but also were able to trap water soluble metal ions. Reduction of trapped metal ions in the polymeric shell of MWCNT-g-PCA hybrid materials by reducing agents such as sodium borohydride led to encapsulated metal nanoparticles on the surface of MWCNT. Herein palladium nanoparticles were encapsulated and transported by MWCNT-g-PCA hybrid materials (MWCNT-g-PCA-EPN) and their application as nanocatalyst toward... 

Sulfur quantum dots (SQDs) were fabricated using a facile hydrothermal method and used for the preparation of functional food packaging film and compared the properties with other sulfur-based fillers like elemental sulfur (ES) and sulfur nanoparticles (SNP). The SQDs have an average size of 5.3 nm and were very stable in aqueous suspension. Unlike other sulfur-based fillers, the SQD showed high antioxidant, antibacterial and antifungal activity, but no cytotoxicity was found for L929 mouse fibroblasts even after long-term exposure of 48 h. When sulfur-based fillers were added to the alginate film, SQD was more evenly dispersed in the polymer matrix than SNP and ES. The addition of SQD to...