Sharif Digital Repository

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... 

Causal discovery is of utmost importance for agents who must plan and decide based on observations. Since mistaking correlation with causation might lead to un- wanted consequences. The gold standard to discover causal relation is to perform experiments. However, experiments are in many cases expensive, unethical or impossible to perform. In these situations, there is a need for observational causal discovery. Causal discovery in the observational data setting involves making significant assumptions on the data and on the underlying causal model. This thesis aims to alleviate some of the assumptions and tries to identify the causal relationships and causal mechanisms using generative neural... 

In recent years, deep generative models have achieved incredible successes in various fields, including graph generation. Due to the advances made in graph generation by deep generative models, these methods have shown numerous applications from drug discovery and molecular graph generation to modeling social and citation network graphs. Graph generation is an approach to discovering and exploring new graph structures and has been attracting growing attention. One of the most challenging applications of deep graph generative models is molecular graph generation since it requires not only generating chemically valid molecular structures but also optimizing their chemical properties in the... 

Manual code review, essential for software quality, suffers from time constraints and repetitive tasks. This thesis investigates using large language models (LLMs) with prompt engineering and in-context learning to automate aspects of the process. By leveraging an LLM's generalization capabilities, we aim to achieve automation with limited resources and minimal pre-training. Due to the development of large language models and in-context learning capabilities, our proposed approach is to add contextual information relevant to code review to the model input. Our approach focuses on providing context-specific samples and documents related to the reviewed code, enabling the LLM to learn from the... 

Shadows are one of the most important details in graphical images. There exist many shadow generation algorithms each of which suffers from some problems. High processing time and ill-shaped shadow borders, known as alasing, are some of such problems. In this paper, we propose a heuristic method based on standard shadow map technique, named rotated shadow maps, to generate excellent hard shadows. Our method uses some shadow maps which are generated from the same view but objects are slightly rotated around the center of view area. Rotated shadow maps can be considered as an independent hard shadow generation algorithm. Also it can be combined with other shadow map based approaches. Utilizing... 

Recently, a new generation of machine learning tasks, namely data generation, was born by emerging deep networks and modern methods for training neural networks on one hand, and the growth of available training data for training these networks on the other hand. Although distribution estimation and sampling were well-known problems in the science of statics, deep generative models can properly generate samples from real world distributions that common statistical methods fail in them e.g., image and music generation.Due to these improvements in deep generative models, researchers have recently tried to propose deep generative models for datasets with complex structures. These structured... 

A large number of machine learning problems are considered as structured output problems in which the goal is to find the mapping function between an input vector to a number of variables in the output side which are statistically correlated. Motivated by the advantages of simultaneous learning of these variables compared to learning them separately, many structured output models have been introduced. Decreasing the sample complexity, increasing the generalization ability and overcoming to noisy data are some of these benefits. So in the first step of this research we concentrate on one of classical but important problems in bioinformatics which is automatic protein function prediction.... 

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... 

There have been many works on generative networks and image generation in the past few years, but the problem with this work is that there is no control over the generated images. The goal of disentangled image synthesis is to generate new images with specific detail and have control over the generated images. Image-based virtual try-on aims to synthesize the customer image with an in-shop clothes image to acquire seamless and natural try-on results, which have attracted increasing attention. The main procedures of image-based virtual try-on usually consist of clothes image generation and try-on image synthesis. In contrast, prior arts cannot guarantee satisfying clothes results when facing... 

Representation learning refers to mapping the input data to another space, usually with lower dimensions than the input space. This task can be helpful in improving the performance of methods in downstream tasks, compression, and improving sample generation in generative models. Representation learning is a problem connected to information theory, and information theory's concepts and quantities are used widely in representation learning models. Besides, the representation learning problem is closely related to latent variable generative models. These models usually learn useful representations in their process of training, implicitly or explicitly. So, the usage of latent variable... 

Distributed generation (DG) technologies have achieved an intense progress during the last years. One of the appropriate ways to enter a DG system in the electrical network is through the use of microgrids. Because the majority of DG units are connected to the grid with power electronic converters, specific control strategies have been developed for these converters. Control strategies of DGs in microgrid differ in the grid-connected and the islanded states of operation. Changing the control modes from one state to another requires a reliable islanding detection method, and results in unwanted transients. It is, therefore, desirable to devise an integrated control approach for both states.... 

Human walking is widely recognized as one of the most adaptable and robust forms of locomotion in nature, with intricate neural and biomechanical systems interacting to support this complex behavior. It is proposed that these systems are organized in a hierarchical structure, with the lower level comprising a complex distributed system consisting of muscles and the spinal cord, and the higher level being the brain cortex. The higher level is responsible for training and monitoring the output of the lower level, and intervening when the lower system fails to stabilize the system. To control the lower level, one popular model that has emerged is the central pattern generator (CPG). It is... 

The purpose of this project is designing a controller for a floating offshore wind turbine to exctract maximum power from wind energy in the first operating region of wind turbine. The NREL 5 MW OC3-Hywind floating wind turbine with BARGE platform will be used for this project. By using Turbsim software, proper wind profile for the first operating region of wind turbine will be produced. This profile will be excecuted for the FAST software that is simulates the mechanical part of wind turbine. By implementing the doubly fed induction generator in Simulink enviroment and designing the controller for doubly fed induction generator, the desired torque for maximizing the exctraction power of... 

A microgrid system consists of some distributed generators (DGs) and local loads which can operate in both grid-connected and islanded modes of operation. An islanded mode of operation is required when the grid voltage deteriorates and its power quality drops. This islanding operation can be attained by opening grid-side switches or by using back-to-back (b-to-b) converters. Due to the increase of highly sensitive loads and to have better power control and higher power quality, the use of back-to-back converters has recently attracted interests among researchers. However, the major drawback of this solution is the cost of a b-to-b converter. Therefore the aim of this project is to gain some... 

The use of distributed generation (DG) units provides several advantages for the utility distribution grid. Mostly power electronics converter is used as an interface to connect a DG unit to a utility grid. A DG unit normally operates in a grid-connected mode. In this mode, the host grid dominantly dictates the voltage amplitude and frequency of the local load at the point of common coupling (PCC) and the DG unit controls its real/reactive power components. Separation of the DG unit and its dedicated load from the host grid is called islanding. Subsequent to the formation of an island, due to the lack of control over the voltage magnitude and frequency, the islanded DG system becomes... 

Load frequency control (LFC) is one of the important control problems in interconnected power system design and operation, and is becoming more significant today because of the increasing size, changing structure, emerging new uncertainties, environmental constraints and the omplexity of power systems. Therefore, load frequency function requires increased performance and flexibility to ensure that it is capable of maintaining a generation-load balance, following serious disturbances.Real power systems are often large-scale systems which are composed of many interacting subsystems. Therefore, it is difficult to control such systems due to the required inherent computational complexities and... 

Concerns about fossil fuels and the consequent effects of burning such fuels on the environment have drawn attention of different countries around the world to the development of renewable energy systems, and it has encouraged researchers to focus on expanding the methods of using different sources of renewable energy. One of the most important renewable energies is wind energy, which is generated by wind turbines. A cluster of wind turbines forms a wind farm. Wind turbines can be equipped with a supervisory control system that sends the data measured by the sensors to the operator. Design of supervisory control algorithm for a wind farm requires a model of the wind farm system. The task of... 

Three-phase voltage source converter (VSC) is vastly utilized in the renewable energy resource and distributed generation (DG) systems interface units due to its high controllability and flexibility. Because of the high nonlinear nature of the system, the gird-connected inverter shows poor performance if the conventional control scheme are used. In this thesis, a new control scheme based on backstepping strategy is presented for control of a grid connected DG systems. The proposed scheme is responsible for regulating the active and reactive power components of the VSC power of the DG system. The VSC of the DG system is connected to a stiff utility grid through an L-type filter. In the first... 

Nowadays, the use of distributed generation (DG) units and microgrid systems is mainly limited to their operation in the grid-connected mode. In the grid-connected mode, a microgrid including its loads and DG units is connected to the main grid at the point of common coupling (PCC). In this case, the main grid provides fixed voltage and frequency for the microgird and each DG unit regulates its power components based on the conventional dqcurrent control strategy. Autonomous (islanded) operation of a microgrid, however, is not as straightforward as its operation in the grid-connected mode, i.e., control and power management of an islanded microgrid is very difficult due to the stochastic...