Sharif Digital Repository

Applying of seismic waves and analysing of its reflections is of effective ways in exploration of the internal structur of the Earth. In this method the Earth is considered as a elastic body that seismic waves are propagated in it. In this thesis the how of creation and propagation of seismic waves is modeld at first and then the inverse problem, exploration of the internal structure of the Earth from observed data, is considered  

Visual odometry(VO) is the process of estimating the egomotion of an agent(e.g., vehicle, human, and robot) using the input of a single or multiple cameras attached to it. Application domains include robotics, wearable computing, augmented reality, and automotive. The term was chosen for its similarity to wheel odometry, which incrementally estimates the motion of a vehicle by integrating the number of turns of its wheels over time. Likewise, VO operates by incrementally estimating the pose of the vehicle through examination of the changes that movements induces on the images of its onboard cameras. For the VO to work effectively, there should be sufficient illumination in the environment... 

We show that the equations of reinforcement learning and light transport simulation (rendering equation) are related integral equations. Based on this correspondence, a scheme to learn importance while sampling path space is derived. The new approach is demonstrated in a consistent ray tracing algorithm that uses reinforcement learning to progressively learn where light comes from. As using this information for importance sampling includes information about visibility, too, the number of light transport paths with non-zero contribution is dramatically increased, resulting in much less noisy images within a fixed time budget  

Due to immense increase in availability of text data, interest in using machine learning models to solve problems previously impossibly costly has increased significantly. The first step is to represent natural language in a form that is easy for the machine learning algorithms to work on. Recent advances in learned representation of text data using simple neural networks(e.g. word2vec and doc2vec) helped increase performance of natural language processing on downstream tasks. Here we show that methods like doc2vec that were examined mostly in the English language can be used on Persian(Farsi) with little modification. To Demonstrate this, we use text classification tasks, and train... 

Optimal drug dosing is one of the fundamental challenges in treating complex diseases such as cancer. Cancer, characterized by the rapid growth of tumors and the stimulation of angiogenesis to form new blood vessels for providing nutrients and oxygen, is one of the most significant public health threats. Angiogenesis is influenced by factors that guide the migration and aggregation of endothelial cells toward the tumor. In this study, the Anderson-Chaplain mathematical model was utilized to simulate the dynamics of angiogenesis and analyze the effects of relevant biological parameters.To better understand this process, topological data analysis and clustering methods were applied to the... 

In the past, opinion polls have generally been conducted using statistical techniques and the comments of those samples, which, in addition to its various challenges, require much power and cost per poll. Today, due to the high penetration rate of social networks in society, a very large fraction of people discuss such issues in terms of their opinions and preferences in these networks, and hence, these networks can be a valuable source of information to get the opinions of people in a short time, and by lower cost.The purpose of this thesis is to enhancing the results of the surveys which have already performed using various methods of text processing in social networks by using the graph... 

One of the important tasks in machine vision is 3D structure reconstruction or depth estimation. In recent decades, methods based on machine learning have been presented, which can be used to perform this costly task more optimally. For this purpose, two branches of supervised and unsupervised learning have been considered. In supervised learning, the available data is in the form of image-depth map pairs, such as the kitti dataset. The image is fed to the network and the output is made in the form of a depth map. The cost function of this network is obtained by comparing the generated depth map and the depth map in the data set calculated by laser. In unsupervised learning, there is no... 

The numerical solution of differential equations in high dimensions has always been a challenge and has been associated with various computational difficulties. These equations appear naturally in a variety of problems such as financial mathematics, control, and physics, and their optimal solution with high accuracy and speed can open new windows on new applications. Conventional methods such as Finite element and finite difference method in high dimensions lose their efficiency, which is a barrier to fast and accurate calculation of these equations. In this dissertation first, we review some theoretical and practical aspects of deep neural networks and then we try to examine the recent... 

In this thesis which is an extensive expository and survey, we study concepts like irreducible, primitive, regular and almost regularournaments. We study and show the proof of some graph theoretical properties of tournaments such as being pancyclic, vertex or arc-pancyclic, and a lower bound on the number of cycles of each length in irreducible tournaments. Spectral properties and the Perron eigenvalue of tournament matrices are also extensively discussed.The ranks of tournament matrices over different algebraic fields, is another subjectthat is covered. We have written computer programs to study the distribution of the rank of tournament matrices over Z2.With a little more work,one might... 

With the increasing advancement of technology and available information, finding suitable content for users is one of the significant challenges. Recommender systems have emerged as an effective solution to confront these challenges. Reinforcement learning can be employed as an effective approach to designing and implementing recommender systems. However, there are various challenges and complexities in the context of reinforcement learning applied to recommendation systems. In this context, Online users are treated as the environment, and concepts like reward functions and environment dynamics are not clearly defined, complicating the reinforcement learning process. In this thesis, a... 

In recommender systems, the available data structure is mostly in the form of a graph; Therefore, during the past years, graph based recommendation methods, especially Graph neural networks, have received a lot of attention. These networks obtain the representation of user nodes and items by aggregating multi-hop neighbors in different ways. Unfortunately, the existing methods are part of the Supervised learning scenario and due to the lack of data in recommender systems, they have faced serious challenges. These challenges reduce the ability to learn to representation nodes and the accuracy of recommendations. Recently, knowledge graph based recommendations have achieved excellent... 

Today with the availability of cheap depth sensor, processing point clouds produced by these sensors and extracting geometric features is an active field of computer vision. Object recognition is a basic computer vision issues that even with considerable research has remained as a challenge. In these thesis we have studied methods of utilizing depth images and geometric information of point clouds in order to extract geometric features from point clouds and have introduces a set of new geometric features using Normal Orientation Histogram. Also a novel and efficient method for segmentation of point cloud of indoor scenes is proposed. Experimental results depict that our proposed methods have... 

In this thesis, we study a continuous time financial market of some risky assets and a risk-free asset for investment in a finite time period. We use mean-variance approach for investment in this market. In the model considered here, the mean returns of individual assets are explicitly affected by underlying Gaussian economic factors. Using past and present information of the asset prices, a partial-information stochastic optimal control problem with random coefficients is formulated. Here, the partial information is due to the fact that the economic factors can not be directly observed. In first step, by filtering and in secound step by solving the stochastic control problem, we show that... 

Face recognition, which is one of the most important biometrics, has always been one of the main challenges in many security issues, such as verifying the identity of customers of financial institutions and passengers at the airport, and such issues have many applications in daily life. Face recognition has always been an important issue in computer vision and pattern recognition. Currently, several methods based on deep networks have shown great results in face recognition, among which the following can be mentioned.1.The deep face was introduced by Facebook in 2014; 2.Face-net was presented by Google in 2015 ;3.VGGFace was presented by Oxford University in 2015; 4.Openface was presented by... 

Non-overlapping Cameras in multi-camera systems have become prevalent in robotics and computer vision research; therefore, it is possible to cover the wide field of view, and researches have been done for computing extrinsic parameters of cameras. These cameras do not have any overlap in their views, so obtaining the corresponding point in their images is somehow impossible. Light and shadow geometry is analogous to Structure from Motion problem. In this thesis,we study Structure from Motion problem and have tried to propose an approach for estimating extrinsic parameters of non-overlapping cameras in Multi-camera systems. We formulate the problem by using light and shadow geometry and... 

Tissue engineering is an interdisciplinary field that incorporates principles of engineering with the life sciences. Tissue engineering consists of three principle; cells, scaffolds for cells expansion and growth factors. These factors together can help tissue engineers to provide appropriate microenvironments for directing special cells behavior.
Fibrous scaffolds have found many applications in tissue engineering due to their nanometer dimensions and high surface to volume ratio. Fibrous scaffolds can impart mechanical strength, structure for cell attachment and act as reservoirs for biomolecule delivery in much a same way as the natural fibrous components of the extracellular...