Sharif Digital Repository

Improvements in deep neural networks and their widespread use in research and practical application have raised significant concerns about the robustness of these networks against adversarial examples designed to deceive a deep network in calcu- lating the correct output with a slight change in input. Since this is an essential issue in highly sensitive applications, it is necessary to use a training method that reduces the model’s sensitivity to these changes but still preserves the accuracy. The common method for this goal is training the model with adversarial examples. In other words, adversarial examples are generated during the training, and the model is trained with them. This allows... 

Stochastic block models are the most common statistical models for simulating graphs with block structures. These models have been studied extensively for community detection problem and evaluating the performance of algorithms for community detection. Commu- nity detection has various applications in the study of social networks, protein networks, image processing, and natural language processing. In the classical setting, community detection is studied when only one graph is available. In 2021, a model for correlated stochastic block graphs was introduced, in which multiple edge-correlated graphs with the same block structure are observed. In this model, node labels are not available,... 

Intrusion Detection System (IDS) is an equipment destined to provide computer networks security. In recent years, Machine Learning and Deep Neural Network (DNN) methods have been considered as a way to detect new network attacks. Due to the huge amounts of calculations needed for these methods, there is a need for high performance and parallel or specific processors, such as Application Specific Integrated Circuit (ASIC), Graphical Processor Unit (GPU) and Field-Programmable Gate Array (FPGA). The latter seems more suitable than others due to its higher configurability and lesser power consumption. The goal of this study is the acceleration of a DNN-based IDS on FPGA. In this study, which is... 

The growth of user traffic and appearance of new services, has changed architecture of mobile networks. Data flows have different requirments based on their services. Mobile platform should be able to recognize services and slice network resources based on them. Different approaches have been investigated using relevant standards and numerous articles. The focus of this thesis is on implement a platform based on fourth generation mobile networks that can differentiate between data flows according to their respective required QoS.We have implemented an IMS network and connected it to a mobile core network But IMS network just supports session-based services. We introduced a solution for non... 

Detecting out-of-distribution (OOD) input samples at the inference time is a key element in the trustworthy deployment of intelligent models. While there has been a tremendous improvement in various flavors of OOD detection in recent years, the detection performance under adversarial settings lags far behind the performance in the standard setting. In order to bridge this gap, we introduce RODEO in this paper, a data-centric approach that generates effective outliers for robust OOD detection. More specifically, we first show that targeting the classification of adversarially perturbed in- and out-of-distribution samples through outlier exposure (OE) could be an effective strategy for the... 

The price elasticity of gasoline demand is a key parameter in evaluation of various policy options. However, most of the literature uses aggregate data to identify this elasticity. Temporal and spatial aggregation make elasticity estimates unreliable. We employ a unique dataset of all gasoline transactions in Iran during a four-month period around an unexpected exogenous price change to identify price elasticity. After controlling for date and individual fixed effects we estimate a robust significant price elasticity of -0.127. Aggregation of the same data by week, month, and city yields an estimate of -0.3 indicating a significant bias in earlier studies. We also identify a significant... 

Clustered Regularly Interspaced Short Palindromic Repeats, or in short, CRISPR is a relatively new technology that enables geneticists and medical researchers to edit parts of the genome by removing, adding, or altering parts of the DNA. Initially found in the genomes of prokaryotic organisms such as bacteria and archaea, this technology can cure many illnesses such as blindness and cancer. A significant issue for a practical application of CRISPR systems is accurately predicting the single guide RNA (sgRNA) on-target efficacy and off-target sensitivity. While some methods classify these designs, most algorithms are on separate data with different genes and cells. The lack of... 

Operating systems employ free space of main memory as a buffer cache to reduce the number of accesses to the disk subsystem with the aim of improving performance. Main memory which consists of DRAM modules, suffers from data loss in case of power failure and also high cost per gigabyte. Therefore, buffer cache layer needs to redirect critical writes to the disk subsystem which degrades the performance. Non-Volatile Memories (NVMs) are emerging memory devices that can preserve data without need for power current. NVMs, however, cannot replace DRAM due to the shortcomings such as limited endurance and low performance. Previous studies suggested employing hybrid DRAM-NVM buffer caches to... 

Deteriorated old urban neighborhoods usually have no access to appropriate municipal services, and have small-sized buildings, densely populated areas, and earthquake-prone structures. Residents in these areas do not have the will and financial ability to renovate their aged properties. A questionnaire survey of the citizens living in the deteriorated areas of the City of Tehran showed that 80% of households do not have a positive monthly financial balance and cannot afford the cost of reconstructing their properties. Serious drawbacks of previous attempts to renovate deteriorated areas, including lack of adequate government funding, and adding to the dependence of local citizens on... 

Expansions of large urban areas, as well as increase in the rate of car ownership, has given rise to the urban traffic congestion. One of the major approach to cope with the congestion issue is named Transit Oriented Development (TOD), which will cause changes in trip departure times, route choice, mode choice, and trip destinations. TOD also affects the land use in long-run periods. The goal of this study is to clarify the long-term impacts of TOD on land use and find the appropriate transit projects in order to abate the drawbacks of changes in land us. Land use/ transportation has been the subject of many recent studies. One of the previous studies investigates land use and transportation... 

There is a two-way relationship between land use and transportation. So each of them can be used to manage and control the other one. Land use can be Auto-Oriented, or Transit-Oriented. Auto-Oriented Development leads to urban sprawl, causing longer trips and increasing use of private cars. Problems related to this type of development appeared by the end of the 20th century. On the other hand, Transit-Oriented Development may lead to higher density areas, and higher transit use, providing better space for pedestrians and cyclists, as well as increasing environmental pollutions. Land use-transportation relationship hs been the subject of many recent studies. One of the previous models... 

Pavement maintenance and rehabilitation is an important problem in road pavement programming. Pavements experience significant distress over its life due to passage of vehicle (particularly heavy trucks), as well as abrupt variation in temperature (freeze and thaw). This would deteriorate pavements and reduces their serviceability. Rise of pavement maintenance costs places higher importance upon the need for maintenance programming. This study employs an evolutionary model to optimize the long-run benefit of the maintenance program. The model is stochastic in nature, and imbeds a Markovian decision procedure in the model to take into account such variabilities. The model is applied on... 

Conversation modeling is one of the most important goals in the field of understanding natural language and machine intelligence. Recently, with the enormous growth of the Internet and social networks, the amount of available data on the Web has increased significantly.This makes it possible to use data-driven approaches to solve the modeling problem of conversation.One of the most recent data-driven methods is the sequence to sequence modeling. In this document, after providing the necessary prerequisites, we examined the various models that have used the sequence to sequence approach for conversation modeling. We further examined the ways of improving the efficiency of this modeling... 

Wireless sensor networks (WSNs) are tiny devices with limited computation and power supply. For such devices, data transmission is a very energy-consuming operation. Data aggregation eliminates redundancy and minimizes the number of transmissions in order to save energy. This research explores the efficiency of data aggregation by focusing on different aspects of the problem such as energy efficiency, latency and accuracy. To achieve this goal, we first investigate data aggregation efficiency with constraint on delay, which can be compatible with other important system properties such as energy consumption and accuracy. By simulation, we will show that, depending on the application, we can... 

The purpose of this study is to solve the problem of allocating relief teams to reopen damaged roads and remove debris from damaged areas after the earthquake for large cities. In the deteriorated areas of large metropolises, severe earthquakes may damage links, which causes some nodes to separate from the network. In this research, by using an Ant System meta-heuristic algorithm, a joint problem of damaged link restoration and demolished zone debries removal problem for search and rescue of the trapped population in an earthquake is solved. Ant System parameters are calibrated, and sensitivity analyses are done for the number of relief teams, earthquake intensity. The results of 1000... 

Speech recognition is one of the first goals of speech processing. Our goal in this thesis is to use deep learning for speech recognition. In recent years little improvement of speech recognition accuracies are reported. Deep learning is a new learning algorithm that results in improvement in many machine learning tasks. Following improvements reported in speech recognition in English language by deep learning, in this thesis we tried to improve accuracy over common and new recognition methods for Persian language.
First the overall structure of a typical speech recognition system is introduced. For this purpose, the modules of a speech recognition system are introduced. Deep multilayer... 

Design and production of a drug is a very time and money consuming process. It takes more than a decade and about 2.5 million dollars on various stages to design a drug. Attempts to reduce this cost and time to market will make drugs available to customers at a more reasonable time. Some stages such as animal testing phase and clinical trials, can not be replaced and must take place in practice. Fortunately, some laboratory steps are interchangeable with software algorithms. These algorithms can significantly reduce the cost and time to market of the drug if they are accurate enough. On the other hand, the remarkable results of machine learning, in particular, Deep Neural Networks, in areas... 

Applying spoken dialogue systems (SDS's) is growing in the real life more rapidly because of the advances in the design and management of these systems. The traditional touch tone computer telephony systems are being substituted by the SDS's. In a typical SDS, the user speaks naturally to the system through a phone line and the system provides the required information or performs the required action. Banking and ticket reservation are typical examples of the prevalent SDS's. A spoken dialogue system has four units: automatic speech recognition (ASR), natural language understanding (NLU), dialogue management (DM), and spoken language generation (SLG). In this work, the first spoken dialogue... 

In this study, we aim to learn representations from microscopic cell images that effectively capture the features of drugs affecting the cells, allowing us to identify effective drugs for treating a disease. We employ two parallel learning paths using predictive and generative models. Specifically, we have achieved a predictive model on the RxRx19a dataset that, unlike previous models, is interpretable, optimized, and robust to dras- tic changes in drug properties. Additionally, we have developed the first generative model on this dataset, which not only generates high-quality images but also discovers a meaningful latent space. This latent space divides the representation into relevant... 

Road network expansion is a well known problem in the world, particularly in developing countries. One approach in financing this development is to charge the users for the services rendered. Recent studies prove the merits of private sector participation in road investment and revenue collection. Public sector may cooprate with private sector in permitting it to build the roads, operate and maintain them for certain period of time, while taking advantage of the revenues collected. The road will, then, be transferred to the public sector when that period of time is over. This study presents an extension to a previous model in this area. It presents a bi-level programming network design...