Sharif Digital Repository

In traditional machine learning approaches to classification, one uses only a labeled set to train the classifier. Labeled instances however are often difficult, expensive, or time consuming to obtain, as they require the efforts of experienced human annotators. Meanwhile unlabeled data may be relatively easy to collect, but there has been few ways to use them. Semi-supervised learning addresses this problem by using large amount of unlabeled data, together with the labeled data, to build better classifiers. Because semi-supervised learning requires less human effort and gives higher accuracy.Formally, this intuition corresponds to estimating a label function f on the graph so that it... 

The purpose of this research is to monitor the health of structures using the updated finite element model, in which digital images are used to optimize the numerical model. Structural Health Monitoring (SHM) is always an important and significant issue that has attracted the attention of many researchers in recent years. In general, some researches have been conducted in this field using physical sensors that provide discrete data to the system for analysis. Using cameras to monitor the structure makes it possible to extract continuous and integrated data from the structure using digital images, which is a significant advantage compared to physical sensors.In this research, a steel... 

In this study, the aerodynamic flow behavior in an low Reynolds number regime (Re = 1e3) around a tandem configuration is numerically investigated. In such conditions, the flow exhibits characteristics such as laminar behavior, separation bubbles, and high sensitivity to geometric parameters. Initially, using the finite volume method and solving the steady-state Navier–Stokes equations, the flow field around two consecutive airfoils was simulated. The aim of this step was to quantitatively and qualitatively examine the effects of spacing and angle of attack on flow characteristics, including pressure distribution and drag/lift coefficients. Subsequently, the numerical data were used to train... 

Aluminum Bronzes have many applications in marine environments. These alloys suffer from both hot cracking and cold cracking. In order to overcome the hot cracking and cold cracking, Semisolid Stir Joining method and a proper thermal cycle was used respectively. Effects of temperature, stirring rate, and tool type were investigated in Semisolid Stir joining method. In this method, butt joint design was used in order to place specimens, and the specimens were heated up to specific temperatures (920, 925, 930°C). A stirrer (Cylindrical and Grooved tool) with three rotational speeds (800, 1200, 1600 RPM) was introduced into the stir weld seam. Welded specimens were cooled to the 900°C... 

Nowadays, one of the problems of network designers and users is the correctness of network equipment functionality. Since routers play an important role in information transfer between various networks and communicate with each other through a routing protocol, any violation of such protocol can result in the network's failure or defective transfer of data. Thus, being assured of the conformance of the implemented protocol in a router with what has been defined in the protocol's standard is of great significance. To this end, before using or exposing a router, network equipment testers should run protocol conformance tests on it and insure its correct functionality. This brings up the... 

In order to increase the network performance, to discover network security threats, to guarantee the quality of services, to simulate networks and develop new network protocols, network designers and operators need software tools to analyze the network traffic. The first step to do this analysis is network monitoring. This can be done by using various software and hardware tools.
In this research we developed a tool which analyzes the network traffic using appropriate statistical methods. The models which are analyzed in this application are: Poisson, Self-similar and Weibull. This application analyzes the network traffic based on this statistical models and calculates their parameters.... 

The increasing number of road accidents and fatalities have yielded the introduction of several solutions by related industries such as automobile manufacture. Proposed solutions can be classified into two categories: passive, such as airbags, and active such as ABS ad ITS-based technologies. Inter vehicles communication let the communication between vehicles(V2V) or between vehicles and road infrastructure (V2I) by relying on DSRC technology and standards. Intelligent Transportation System (ITS) has three main types of applications namely safety, traffic and commercial. Safety applications are the main goal of vehicular communication, and real time guarantee in safety applications is so... 

In real-time systems, jobs that miss their deadline are considered worthless or ineffective. Nowadays, real-time applications (such as multimedia) that communicate through software-defined networks (SDN) are widely used and will continue to expand in the future. SDN is an emerging network architecture in which control plane is separated from data plane to manage it directly. In this research, we make changes to the controller and the switch used in such network to deliver real time packets or real-time flows (such as video) on time, as much as possible. As missed deadline packets are worthless and they waste bandwidth, dropping them is a solution. Moreover, the controller can drop flows that... 

Network Function Virtualization (NFV) is a new paradigm in computer networks for implementing network functions on general-purpose common-of-the-shelf processors. The revolutionized NFV architecture has shown to be a suitable replacement for hardware middleboxes. NFV is in high demand because of its lower costs, increased flexibility, and more scalability. However, the ever-growing volume of network traffic and increased data transfer speed in computer networks have made achieving high performance the major issue in this area. With thousands of processing cores and high memory bandwidth, Graphical Processing Units (GPUs) are potentially an appropriate solution to accelerate network... 

This thesis describes different neural networks models and an approximation based neural network controller for autonomous underwater vehicles (AUVs). The online multilayer perceptron neural networks (OMLPNN) have been designed to perform modeling of AUVs of which the dynamics are highly nonlinear and time varying. The online recurrent multilayer perceptron neural networks (ORMLPNN) have been additionally designed to generate a memory to pervious states and increase the performance of the modeling. The designed OMLPNN and ORMLPNN with the use of backpropagation learning algorithm have advantages and robustness to model the highly nonlinear functions. The proposed neural networks architecture... 

Advanced Driver Assistance System (ADAS) is a major area of research in today’s vehicle control and safety. In General ADAS can be seen as a mother system with various subsystems to help the driver from pre-crash to post-crash situations. One of the subsystems for ADAS is the Collision Avoidance System (CAS). Nowadays, many researchers are working on different Aspects of CAS, from perception to decision making. Many challenges exist while developing a CAS, but today the greatest challenge is decision making of CAS and its interaction with driver’s behavior. Driving is a complex task and since driver’s consciousness, gender, age and reflexes can cause different behaviors; it is, therefore,... 

Nowadays, the significance of securing data and information is undeniable. Cryptography is being used to provide data security. In addition, cryptanalysis is required to evaluate the effectiveness of cryptography methods, and hence, it is an essential concept for securing data. In general, the cryptography functions shall be designed in a way to impose a high load of time-intensive operations to prevent an adversary from accessing the main data from the encrypted data. As a result, cryptography and cryptanalysis algorithms need high performance computations. So far, a number of methods have been proposed to support the required performance. These methods include: distributed computing and... 

Congestion is an important issue in the network environment. To keep stable the perfor-mance of the network, congestion control algorithms have been extensively studied. Queue management method employed by the routers is one of the important issues in the congestion control study. Active Queue Management (AQM) has been proposed for early detection of congestion inside the network. AQM mechanisms control the queue length in a router by dropping arriving packets. The Random Early Detection (RED) is the most popular AQM mechanism used in routers on the Intenet to allow network users to simultaneously achieve high throughput and low average delay. The RED algorithm may cause heavy oscillation of... 

Nowadays, information gathering based on wireless sensor network (WSN) has formed various applications and it also seems to be spreading more in future. Wireless sensor network, is a network composed of multiple nodes, without any infrastructure, each having one or more sensors, wireless communications unit, processing unit, memory, and power supply. Memory capacity, computation power, and power supply are usually considered as valuable resources and limited in nodes. Most of sensor nodes are located in a dense mode and they are susceptible to failure. Therefore network topology may vary dynamically. A sensor network can support real-time communications independent of its type of structure...