Sharif Digital Repository

Nowadays, video streaming over peer-to-peer networks has become an interesting field to deliver video in large scale networks. As multi-channel live video streaming networks increase,distributing video with high quality among channels faces many challenges. The most significant challenges cause from frequent channel churns, unbalanced channel resources, network heterogeneity and diversity of users’ bandwidths. They include: nodes’ unstability, low users participations, large startup and playback delays, low video quality received by users and lack of resources in unpopular channels.In order to solve the above problems, we have proposed several solutions such as: 1- using distribution groups... 

Semi-supervised Learning (SSL) is an extremely useful approach in many applications where unlabeled data can be easily obtained. Graph based methods are among the most studied branches in SSL. Since neighborhood graph is a key component in these methods, we focus on methods of graph construction in this project. Graph construction methods based on Euclidean distance have the common problem of creating shortcut edges. Shortcut edges refer to the edges which connect two nearby points that are far apart on the manifold. Specifically, we show both in theory and practice that using geodesic distance for selecting and weighting edges results in more appropriate neighborhood graphs. We propose an... 

During the last decade, there have been a great number of researches on complex networks.Data aggregation is the first step in the analysis of these networks. However, due to the large scale of them, almost never is there complete information about a network’s different aspects. Therefore, analysis of a complex network is usually done based on the incomplete data. Al-though a good sampling approach in a way that the achieved sample is a good representative of the whole network has its own challenges, analysis of incomplete data causes a significant alternation in the estimation results. Consequently, one of the first problems emerging after sampling is the possibility of predicting the... 

The fast growth of social networks and their wide range of applications have made the anal-ysis of them an interesting field of research. The growth of concern in modeling large social networksand investigation of their structural features leads studies towards community detec-tion in such networks. In recent years, a great amount of effort has been done for introducing community detection algorithms, many of which are based on optimization of a global cri-terion which needs network’s topology. However, because of big size of most of the social networks , accessing their global information tends to be impossible. Hence, local commu-nity detection algorithms have been introduced. In this... 

A great deal of information are continuously generated by users in different contexts such as social networks and online service providers in terms of temporal marked events. These events indicate that what happened to who by when and where.Modeling such events and predicting future ones has interesting applications in different domains such as item recommendation in online service providers and trending topic prediction in online social networks. However, complex longitudinal dependencies among such events makes the prediction task challenging. Moreover, nonstationarity of generative model of events and large size of events, makes the modeling and learning the models challenging.In this... 

Nowadays many of us spend a big part of our daily times on social media.One of the most important research problems in social media analysis is how to engage users. The trace of user activity over these websites is a valuable resource for user understanding and engagement, but this data is very huge and unstructured. An approach to deal with this problem is user behavior modeling. In this process, first a behavioral model is considered for users, then using the activity data and the behavioral model, some parameters are learned. Finally, using the learned parameters, a user profile is constructed for each user. This profile can be used for user engagement and many other applications.... 

The classification of different tumor types is of great importance in cancer diagnosis and drug discovery. However, most previous cancer classification studies are clinical based and have limited diagnostic ability. Cancer classification using gene expression data is known to contain the keys for addressing the fundamental problems relating to cancer diagnosis. The recent advent of DNA microarray technique has made simultaneous monitoring of thousands of gene expressions possible. With this abundance of gene expression data, researchers have started to explore the possibilities of cancer classification using gene expression data and quite a number of Pattern Recognition approaches have been... 

The accessibility of broadband Internet to home users has pushed video broadcasting applications up, in the list of most favorite applications in todays Internet. Such applications involve delivering the data which generated in a single sender to a set of users which are scattered around the world. While it was formerly believed that the IP-Multicasting is the mean to provide this kind of services, the trend has altered to the use of Application Layer Multicasting techniques due to their deployment simplicty. Since their advent, Peer to Peer applications has mitigated many problems previously unsolved in the field of communication. File sharing, distributed file systems, distributed data... 

Error concealment is one of the effective ways to alleviate the effect of packet loss in video communication over error-prone environments. In order to estimate lost macro-blocks, we have employed Bayesian estimation as an efficient and robust framework. Gaussian process regression has been used as the modeling approach through this framework. Considering luminance component as Gaussian process,a minimum mean squared error estimation of the lost macro-block is obtained. This estimator, as a function of the existing data, is only determined by the covariance matrix defined over them. Therefore,the main step in Gaussian process regression, is construction of the convenient covariance matrix... 

In this research, we have used machine learning methods to approach the problem of human activity recognition. As the process of labeling the data in this problem is so costly and time consuming, and regarding the copious available unlabeled data, semi supervised methods have a high performance in this problem. In recent years, graph based methods have became very populaer among semi supervised learning methods. However, constructing a graph on the data which presents their structure in a proper manner has remained a main challenge in these methods. One of the causes of this problem is the existance of the shortcut edges. In this report, we will first introduce a method to solve the problem... 

In this research, we introduce a semi-supervised manifold regularization framework for hu- man pose estimation. Here we aim the three major challenges in discriminative human pose estimation. We utilize the unlabeled data to reduce the need to labeled data and compen- sate for the complexities in the input space. We model the underlying manifold by a nearest neighbor graph. Due to depth ambiguity which is the main challenge in this problem, the true underlying manifold of the data bends and gets too close to itself is some areas which results in poor graph construction. To solve this problem, we argue that the optimal graph is a subgraph of the k-nearest neighbor graph and employ an... 

Mobile Ad hoc networks or MANETs consist of autonomous selfish nodes, which usually have a limited source of energy and choose their own transmission power independently. Nodes are mobile; hence the network topology changes continuously in time, therefore a topology control protocol is required to preserve the network connectivity and to adjust the transmission range of the nodes. Also in the absence of any infrastructure in the network, since the energy source of each node is limited and nodes are not eager to consume it for others, in order to stimulate the selfish nodes to cooperate to provide network services (such as relaying data packets) , we need to exert an incentive mechanism in... 

Communication among people over the emerging networks has been the focus of attention in different branches of science during last decades. Online Social Networks (OSNs), with more than hundreds of millions of users are powerful means for directing information within and across societies. Thus, studying various aspects of OSNs is an important issue for researchers. Due to large number of users and friendship relationships among them, gathering complete information from an OSN is not feasible. On the other hand, hiding users information and crawlers limitations are challenges for gathering complete data. A Common solution for this problem is Sampling from OSNs. Sampling from OSNs (and... 

In this thesis, we have discussed image classification by sparse representation. Sparse representation is used in two different ways for image classification. The first goal of sparse representation is to make an efficient classifier, that can learn the subspace, in which the data lies. In this field we have surveyed various methods. We also proposed a method, called ”Locality Preserving Dictionary Learning” that works approximately better than state of the art similar methods, specially when training data is limited. We have reported the result of lassification on four datasets including MNIST, USPS, COIL2 and ISOLET. Another use of sparse representation, is to extract local features from... 

Compressive Sensing (CS) is a new paradigm in signal processing and information theory, which proposes to sample and compress sparse signals simultaneously and has drawn much attention in recent years. Many signals in lots of applications have a sparse representation in some bases, so CS is used as an efficient way of data compression in many applications such as image processing and medical applications in the last couple of years. Since some of the distributed information in complex networks are compressible too, CS can be used in order to gather the distribted information on the nodes or links efficiently. Traffic analysis and performance monitoring in computer networks, topology... 

Visual tracking is a classic problem, but is continuously an active area of research, in computer vision. Given a bounding box defining the object of interest (target) in the first frame of a video sequence, the goal of a general tracker is to determine the ob-ject’s bounding box in subsequent frames. Utilizing sparse representation, we propose a robust tracking algorithm to handle challenges such as illumination variation, pose change, and occlusion. Object appearance is modeled using a dictionary composed of target patch images contained in previous frames. In each frame, the target is found from a set of candidates via a likelihood measure that is proportional to the sum of the... 

The emergence of applications such as spam detection and online advertising coupled with the dramatic growth of user-generated content has attracted more and more attention to stream classification. The data stream in such applications is large or even unbounded; moreover, the system is often required to respond in an online manner. Furthermore, one of the main challenges of stream classification is that often the process that generates the data is non-stationary. This phenomenon, known as concept drift, poses different challenges to the classification problem.Therefore, an adaptive approach is required that can manage concept drift in an online fashion. This thesis presents a probabilistic... 

Nowadays, Online Social Networks (OSNs) play an important role in the exchange of information among people. Some previous studies indicate that diffusion behavior and network structure are tightly related. Community structure is one of the most important features of OSNs. Access to the whole network topology is the necessary and prevalent requirement for most of community detection methods, so the limited access to full or partial topology can decrease their accuracy. Using traceable information over diffusion network is a solution to surmount this difficulty. In this work, we are concerned with the community detection by only using the diffusion information, while unlike the previous... 

Object tracking is an important tool for video analysis which has been applied to public surveillance, road safety, and scene analysis. In multiobject tracking, the goal is to extract the trajectory of every target using spacial and temporal information. Since new objects can enter the tracking area, an object detector is required to detect their presence. This detector locates target objects in all frames and its (noisy) output is delivered to an associator that returns each object’s trajectory. The associator uses the motion and appearance model of the objects to discover their relation and find the trajectories. In this thesis, each target is partitioned into a number of non-overlapping... 

Hyperspectral imaging is one of the remote sensing methods that has been widely applied in different applications. A hyperspectral image is composed of a set of pixels showing the spectral signatures in different frequency bands recorded by sensor cells. The process that detects the proportion of pure elements in the combination of pixels is called hyperspectral unmixing. Noisy and incomplete data, high mutual coherence of spectral libraries and different sensor settings are some challenges of the unmixing problem. In this work, we focus on semi-supervised linear hyperspectral unmixing in which a spectral library is given. The resulting linear equation is an underdetermind problem with...