Sharif Digital Repository

Governments and organizations need to classify network trac using deep packet inspection systems, by protocols, applications, and user’s behavior, to monitor, control, and enforce law and governance to the online behavior of its citizens and human resources. The high capacity of machine learning in the classication problem has led trac monitoring systems to use machine learning.The development of machine learning-based trac monitoring systems in the eld of research has reached relative maturity and has reached the border of industrial, commercial and governmental use. In the latest trac classi-cation studies using neural networks, as the most ecient machine learning methods, the classication... 

Today, the use of deep learning algorithms in the design of communication systems has received much attention. One of these areas is the partial or total design of these systems using deep networks. The overall design of a communication system using deep networks allows for global optimization and can provide better performance in cases where classical methods have suboptimal performance without significantly increasing the computational load. In this research, a comprehensive architecture for designing communication systems based on Auto-encoder neural networks is presented. This architecture has the same functionality as classical systems, considering all parts of these systems including... 

Graphs are widely used for representing structured data and analysis of them is an important area that appears in a broad domain of applications. Graph processing is of great importance in analyzing and predicting social media users' behavior, examining financial markets, detecting malware programs, and designing recombinant drugs. For example, consider a graph in which nodes and edges show the financial institutions and the financial connection between these institutions, respectively. Financial connection refers to the investment of one institute by another. Based on the graph structure, predicting trade stability and balance is extremely significant in macro decisions.In the last few... 

Software requirements are changing continuously and hence during software evolution and maintenance, source codes changes are being committed in the software repositories. Reading source codes to understand the changes is a very time consuming and tedious activity. Commit messages contain information about code changes that let developers be aware of the essence of the changes without reading the source codes. Unfortunately, due to the pressure of deadlines and lack of time, developers neglect to write these messages. Commit messages can speed up the process of software understanding for developers and also play an important role in software documentation. Therefore, an automated method for... 

Visual Question Answering (VQA) deep-learning systems tend to capture superficial statistical correlations in the training data because of strong language priors and fail to generalize to test data with a significantly different question-answer(QA) distribution. To address this issue, we introduce a Visually Directed Question Encoder to replace the commonly used RNNs in base models. our method uses visual features alongside word embeddings of question words to encode each word. As a result, the model is forced to look at the visual information relevant to each word and it no longer produces answers based on just the question itself. We evaluate our approach on the VQA generalization task... 

Action recognition is one of the important tasks in computer vision that detects the action label in videos that contain only one action. In recent years, action recognition has attracted much attention and researchers have tried to solve it by different approaches.Action recognition by itself does not have many applications in the real world because videos are untrimmed and do not contain only one action. So Temporal Action Localization(TAL) task in which we want to predict the start and end time of each action in addition to the action label has a lot of applications in the real world and for this reason, TAL is a hot research topic. But due to its complexity, researchers have not reached... 

Representation learning methods on graphs have enabled using machine learning methods on graphs' discrete structure by transferring them to a continuous domain. As graphs' structures are not always static and may evolve through time, dynamic representation learning methods have recently gained scholars' attention. Several methods have been proposed to enable the model to update the embeddings graph changes, or new interactions happen between nodes. These online methods could significantly reduce the learning time by refreshing the model as the changes occur, so we don't need to retrain the model with the complete graph information. Moreover, by using the temporal information of interactions,... 

Nowadays millions of images are shared on social media every day , So image inpainting has become an important issue . After advent of Generative adversarial network image inpainting methodes based on deep learning has been revived and significant progress has been made . For a proper image inpainting , The inpainted image must benefit from the appropriate structure and texture in the missing regions . Therefore, in this project , an attempt is made to use a two-stage structure by using Generative adversarial network .in first stage first by using Gabor filters , the image structure is extracted and then the image structure is completed , while the second stage focuses only on the... 

There has been a noticable surge in the usage of machine learning techniques in various fields, such as security related fields. With this growing pace of using machine learning to solve various problems, securing these models against attackers has become one of the main topics of machine learning literatures. Recent work has shown that in an adversarial environment, machine learning models are vulernable, and attackers can create carefully crafted inputs to fool the models. With the advent of deep neural networks, many researchers have used deep neural networks for the task of malware detection, and they have achieved impresive results. Finding the vulnerlabilities of these models is an... 

Detection text in natural images is a challenging task due to the complex backgrounds in an image. complex backgrounds, changes in ambient light, changing viewing angles, and other factors can make systems difficult to detection text. Hence text detection is always an problem. Since detection and recognizing a text in an image has many uses such as translating texts for tourists, helping the blind, etc., recognizing a text in different languages is important. In this thesis, we first examine the three methods of Reading Text in the Wild with Convolutional Neural Networks and FOTS and CRAFT. Then we prepared two Persian data sets. The first data set contains images to which Persian texts have... 

Today, we live in an information age where all people can access the vast amount of data in the world by connecting to the Internet.Since the Internet has expanded significantly to share information, some individuals and organizations seek to be able to prevent the possible sabotage of some people by monitoring network users. Analysis of computer network traffic is one of the importance issues that many activities have been done in this area. One of the most important questions in traffic analysis is to identify the main content of traffic on the encrypted network. Numerous studies have shown that the traffic of websites visited through the Tor network, including Specific information that... 

In this thesis, we review some methods related to model-free discrete reinforcement learning and their corresponding algorithms. Our main goal is to present existing methods in an integrated and formal setup, without compromising their mathematical accuracy or comprehensibility. We have done our best to fix the inconsistencies existing in notations and definitions appearing in different areas of the vast literature. We discuss dynamic programming methods, including policy iteration and value iteration and temporal difference methods as well as policy-based methods such as policy gradient, advantage actor-critic, TRPO, and PPO. Among value-based methods, we discuss Q-learning and C51 where we... 

Cryptocurrencies (cryptos), as a new type of money, are considered a medium of exchange, an investment asset, and a hedging tool in today's world. In 2008, bitcoin as the first cryptocurrency was introduced, which has survived through recent years and has gained more and more popularity every day. Cryptos are one of the first applications of blockchain, the technology that many expect to revolutionize the future world in different ways. We aim to investigate what affects the bitcoin price, based on artificial intelligence and, in particular, machine learning. First, we find features impacting bitcoin price via a thorough investigation of the literature. Then, applying machine learning and... 

Regular monitoring of health factors such as blood pressure and glucose is essential to manage human health. In many such software applications, the patients have to manually enter the value sensed by medical devices such as glucometers into the app. According to medical specialists, this procedure has several drawbacks: (1) Entering values by patients, several times in a day is bothersome, and makes users leave the app, (2) due to the direct intervention of the patient in the procedure, it is error-prone, and besides, (3) users tend to enter unrealistic values. With edge computing, cloud infrastructures, and mobile phones which are ubiquitous and can capture images, it is now possible to... 

Self-supervied learning is a method to reduce the need for large labeled datasets in supervised learning. In self-supervised learning, the goal is to design a pretext task that can be trained without any labels. This pretext task results in learning a representation of data that can reduce the need for labels when used for different tasks. In the domain of images, data augmenting transformations which are often a composition of simple transformations such as random cropping and color jitter have been used for the design of pretext tasks. These simple transformations can cause information loss in some datasets which limits the usage of the learned representations for various downstream tasks.... 

With the fast-growing mobile networks, the need for network reliability increases. Various events like antenna failures, attacks on mobile networks, floods, sporting events cause abnormal behavior in mobile networks. Some events like antenna failures and attacks need consideration. But some of them like sporting events can be ignored. Right now, most mobile operators, detect these events manually which is a very time-consuming and costly task for them. To overcome this problem one idea is to predict future data by prediction algorithms. Then compare new data with the predicted one. If there exists significant deviation, there is an anomaly. But most of the failures and anomalies in mobile... 

Recent success of Deep Learning models, resulted in their evergrowing application in many fields. However these models usually require huge datasets, which can sometimes be hard to collect. One of the challenges related to medical data, is the Batch Effect; Medical data is usually gathered through multiple experiments. Each experiment might have a slightly different conditions than the other, resulting a shift in the data related to that batch. Batch effects can have more severe impact during testing time, as the shift in the data distribution could be bigger. Many methods have been proposed to reduce or remove the effect of external conditions on data distribution.Deep Learning models have... 

DeepFake is a type of forgery that uses deep learning algorithms to make changes to audio and video content that the audience is unable to detect. Nowadays, due to the threats posed by the use of DeepFake to move people's faces in video, researchers' attention has been drawn to designing methods to detect this type of forgery. Detection methods are usually classified into two types. The first case is the extraction of features to detect forgery distortions, for example, the extraction of facial orientations to detect inconsistencies. The second case is the use of deep learning networks for feature extraction and classification, of which the EfficientNet network is an example. Despite the... 

Among all biometric approaches, gait analysis is one of the most practical methods for human identity recognition. Gait has a lot of advantages over other biometrics like face recognition, iris recognition, fingerprint, etc. First and foremost, the gait data can be collected from a distance, and there is no need for subject’s cooperation. Another advantage of this biometric method is its cost-effectiveness and the fact that it does not need high-resolution images. But there are significant challenges in detecting and analyzing this feature. One of the most important challenges is decreased recognition accuracy caused by identity-irrelevant factors like camera viewpoint and changes in walking... 

In this thesis, supervised and unsupervised deep learning-based frameworks were proposed for distributed damage detection and quantification in concrete using sinusoidal ultrasonic response signals. Before the main study on ultrasonic-based concrete damage assessment, a preliminary study was performed on deep learning-based concrete compressive strength prediction. In this study, convolutional neural networks were utilized to predict the compressive strength of concrete through its mix proportions. The Genetic algorithm was employed to find the optimum number of filters in each convolutional layer of the convolutional neural networks. The proposed framework demonstrated high accuracy in...