Sharif Digital Repository

A rotated general regression neural network is presented as an enhancement to the general regression neural network. A variable kernel estimate for multivariate densities is considered. A coordinate transformation is adopted which circumvent the difficulty of predicting multimodal distribution with large variance differences between modes which is associated with the general regression neural network. The proposed technique trains the network in a way that the variance differences between modes is kept small and in the same order. Further, the technique reduces the number of indispensable training parameters to two parameters and lowers the load of the computation as well as the time for... 

This paper introduces a new procedure to forecast the future O/D demand. It is a hybrid of logit and Fratar model. The hybrid model has the long run, policy sensitive, characteristic of a logit model, calibrated at sector-level with little/no zero O/D cells. This feature, joint with a Fratar-type operation at zonal level within a sector, gives a better performance to this model than either of the two types of the models alone. The performance of the hybrid model is contrasted with a neural network model, and shows encouraging results in a real case  

Visual attributes are great means of describing images or scenes, in a way both humans and computers understand. In order to establish a correspondence between images and to be able to compare the strength of each property between images, relative attributes were introduced. However, since their introduction, hand-crafted and engineered features were used to learn increasingly complex models for the problem of relative attributes. This limits the applicability of those methods for more realistic cases. We introduce a deep neural network architecture for the task of relative attribute prediction. A convolutional neural network (ConvNet) is adopted to learn the features by including an... 

Estimation of 3D human pose from a single image is a challenging task because of ambiguities in projection from 3D space to the 2D image plane. A new two-stage deep convolutional neural network-based method is proposed for regressing the distance and angular difference matrices among body joints. Using the angular difference between body joints in addition to the distance between them in articulated objects such as human body can better model the structure of the shapes and increases the modelling capability of the learning method. Experimental results on HumanEva I and Human3.6M datasets show that the proposed method has substantial improvement in the mean per joint position error measure... 

With the recent advancements in deep learning, high performance neural networks have been introduced. These neural networks also can be used to solve similar problems in a transfer learning approach. Recently, several state-of-The-Art Convolutional Neural Networks (CNNs) are proposed for computer vision tasks. On the other hand, in-The-wild No-Reference (Blind) Image Quality Assessment (NR-IQA) problem is known as a challenging human perceptual problem. In this paper, a transfer learning approach is used to solve the problem of in-The-wild NR-IQA. With a few training times, the proposed neural network exceeds all the previous methods which are not using deep neural networks. Further, the... 

Developing time-invariant solutions for recognition of human action is still an important and open challenge. Three issues make time-invariant solutions so important: different speed of performing the same action by different people, latency in doing the actions and the existence of redundant frames in the recorded video. To overcome these problems, we propose a method based on the so-called memory of the joints to remember only the cumulative positive and negative movement of each joint. Hence, we transform action recognition from time-space to shape-space and the action recognition becomes the problem of shape classification. These shape features contain highly discriminative information... 

Neural networks have been used for process fault diagnosis. In this work, the cluster analysis is used to design a structurally optimized Probabilistic Neural Network. This network is called the Clustered-Based Design Probabilistic Neural Network (CBDPNN). The CBDPNN is capable of diagnosing the faults of nonlinear processes operating over several regions. The performance and training status of the proposed CBDPNN is compared to a conventional Multi-Layer Perceptron (MLP) that is trained on the whole operating region. Simulation results indicate that both schemes have the same performance, but, the training of CBDPNN is much easier than the conventional MLP, although it has about 50% more... 

Text processing, as one of the main issues in the field of artificial intelligence, has received a lot of attention in recent decades. Numerous methods and algorithms are proposed to address the task of semantic textual similarity which is one of the sub-branches of text processing. Due to the special features of the Persian language and its non-standard writing system, finding semantic similarity is an even more challenging task in Persian. On the other hand, producing a proper corpus that can be used for training a model for finding semantic similarities, is of great importance. In this study, the main purpose is to propose a method for measuring the semantic similarity between short... 

An experimental based ANN model is constructed to describe the performance of vacuum membrane distillation process for desalination in different operating conditions. The vacuum pressure, feed inlet temperature, concentration of the feed salt aqueous solution, and feed flow rate are the input variables of this process, while response is the permeate flux. The neural network approach is capable for modeling this membrane distillation configuration. The application of Genetic Algorithm to optimize the ANN model parameters was also examined. This is an abstract of a paper presented at the CHISA 2012 - 20th International Congress of Chemical and Process Engineering and PRES 2012 - 15th... 

An experimental based ANN model is constructed to describe the performance of vacuum membrane distillation process for desalination in different operating conditions. The vacuum pressure, the feed inlet temperature, the concentration of the feed salt aqueous solution and the feed flow rate are the input variables of this process, whereas the response is the permeate flux. The neural network approach was found to be capable for modeling this membrane distillation configuration. The application of Genetic Algorithm (GA) to optimize the ANN model parameters was also investigated  

Embryogenesis, regeneration and cell differentiation in microbiological entities are influenced by mechanical forces. Therefore, development of mechanical properties of these materials is important. Neural network technique is a useful method which can be used to obtain cell deformation by the means of force-geometric deformation data or vice versa. Prior to insertion in the needle injection process, deformation and geometry of cell under external point-load is a key element to understand the interaction between cell and needle. In this paper, the goal is the prediction of cell membrane deformation under a certain force and to visually estimate the force of indentation on the membrane from... 

Fin stabilizers are very effective devices for controlling the ship roll motion against external wave-generated moments. Lift forces due to flow around fin with an angle of attack produce anti - roll moment. Therefore control of attack angle plays important role in reducing roll of ships. This paper presents results of using a combined neural network and PID for roll control of ship with small draught. Numerical results are given of around-fin flow analysis with considering free surface effect modelled by neural network and imposed to controlling loop. Hydraulic machinery constraints are also considered in the modelling. The obtained results show good performance of the controller in... 

River water quality monitoring using traditional water sampling and laboratory analyses is expensive and time-consuming. The application of artificial neural network (ANN) models to simulate water quality parameters is cost-effective, quick, and reliable. This study provides two methods of mathematical and ANN modeling to simulate and forecast five important river water quality indicators (DO, TDS, SAR, BOD5, HCO3) correlated with variables such as EC, temperature, and pH which can be measured easily and almost with no cost. The mathematical method is based on polynomial fitting with least square method and the neural network model was developed using a feed-forward algorithm. The 35 years’... 

This paper addresses the design of an observer based adaptive neural controller for a class of strict-feedback nonlinear uncertain systems subject to input delay, saturation and unknown direction. The input delay has been handled using an integral compensator term in the controller design. A neural network observer has been developed to estimate the unmeasured states. In the observer design, the Lipschitz condition has been relaxed. To solve the problem of unknown control directions, the Nussbaum gain function has been applied in the backstepping controller design. “The explosion of complexity” occurred in the traditional backstepping technique has been avoided utilizing the dynamic surface... 

Given the importance of crude oil prices for businesses, governments and policy makers, this paper investigates predictability of oil prices using artificial neural networks taking into account the exhaustible nature of crude oil and impact of monetary policy along with other major drivers of crude oil prices. A multilayer perceptron neural network is developed and trained with historical data from 1980 to 2014 and using mean square error for testing data, optimal number of hidden layer neurons is determined and the designed MLP neural network is used for estimation of the forecasting model. Meanwhile, an economic model for crude oil prices is developed and estimated using a vector... 

Given the importance of crude oil prices for businesses, governments and policy makers, this paper investigates predictability of oil prices using artificial neural networks taking into account the exhaustible nature of crude oil and impact of monetary policy along with other major drivers of crude oil prices. A multilayer perceptron neural network is developed and trained with historical data from 1980 to 2014 and using mean square error for testing data, optimal number of hidden layer neurons is determined and the designed MLP neural network is used for estimation of the forecasting model. Meanwhile, an economic model for crude oil prices is developed and estimated using a vector... 

In this research, we simulate the corrosive behavior of steel reinforcements on 5 different mixtures to investigate the effect of two powerful protective methods, including pozzolanic additives and corrosion inhibitor on concrete, by artificial neural networks (ANNs). Related to this model, fly ash (FA), micro silica (MS), and slag were used as pozzolanic materials at an optimum 25%, 10%, and 25% of cement weight, respectively. Moreover, Ferrogard 901 as an inhibitor was also utilized. The producer recommends using12 kg/m3 to get the best possible results. The non-linear corrosion of concrete into a marine solution (3.5% NaCl) was simulated by the feed forward back propagation (FFBP)... 

In this research, we simulate the corrosive behavior of steel reinforcements on 5 different mixtures to investigate the effect of two powerful protective methods, including pozzolanic additives and corrosion inhibitor on concrete, by artificial neural networks (ANNs). Related to this model, fly ash (FA), micro silica (MS), and slag were used as pozzolanic materials at an optimum 25%, 10%, and 25% of cement weight, respectively. Moreover, Ferrogard 901 as an inhibitor was also utilized. The producer recommends using12 kg/m3 to get the best possible results. The non-linear corrosion of concrete into a marine solution (3.5% NaCl) was simulated by the feed forward back propagation (FFBP)... 

Tourism is one of the largest sources of economic revenue for many countries around the world. The historical and cultural treasures of Iran made it one the main destinations for international tourists. One of the biggest problems encountered by the tourists during the visit to monuments of Iran is the lack of information about the visited landmark. Given that cameras can be found in all of the smart phones, the use of the landmark's photos can be very important for obtaining information about the tourism sites. The detection of the landmarks in an image taken by the mobile phone camera can be a very complex task depending on the angle and the light situation in which the photo is taken. In... 

As recent neural networks are being improved to be more accurate, their model's size is exponentially growing. Thus, a huge number of parameters requires to be loaded and stored from/in memory hierarchy and computed in processors to perform training or inference phase of neural network processing. Increasing the number of parameters causes a big challenge for real-time deployment since the memory bandwidth improvement's trend cannot keep up with models' complexity growing trend. Although some operations in neural networks processing are computational intensive such as convolutional layer computing, computing dense layers face with memory bandwidth bottleneck. To address the issue, the paper...