Sharif Digital Repository

Traditionally, machine learning models for classification tasks rely on statistical methods to find correlations between patterns in the input data and their correspond- ing labels. However, these correlations are not necessarily consistent across different data partitions and may change at test time. Such unstable correlations are referred to as spurious correlations. When the spurious correlation relied upon during training changes at test time, the model’s accuracy can degrade. To improve robustness to shifts in spurious correlations, most research in this area assumes that group annota- tions based on different values of the spurious attribute are available during training or validation.... 

Hyperledger Fabric is an open-source modular platform to launch permissioned-blockchains for use cases like banking and supply chains. This system uses the elliptic curve digital signature algorithm to authenticate transactions. The advent and progress of quantum computers have threatened the security of classical cryptographic schemes; therefore, the process of re- placing these schemes with post-quantum alternatives has gained considerable importance.This research intends to replace the elliptic curve digital signature algorithm with post-quantum digital signatures by using the algorithms passed to the third round of the National Institute of Science and Technologies (NIST) post- quantum... 

This thesis aims to create an electronic health system for transferring, verifying, and monitoring patient data. Blockchain technology is used in this system. In this system, the patient can communicate with the relevant hospital via an authentication protocol and transmit his data to the hospital. The hospital levels the patient's data creates the patient's treatment prescription and shares it with the pharmacy and insurance company to facilitate the patient's effective treatment. Blockchain technology has been used in a variety of electronic health systems. There are several problems with these systems, including the incorrect use of blockchain, a lack of comprehensiveness, a lack of... 

The electronic health system plays a crucial role in modern healthcare, facilitating efficient data management, secure information exchange, and improved patient care. However, existing systems often face challenges related to security, privacy, scalability, and interoperability. The primary objectives of this study are proposing a system that ensures data security, patient and peer privacy, scalability, data integrity, deniability, traceability, secure data sharing, access control, transparency, real-time monitoring, and security in the drug supply chain. To achieve these objectives, we have proposed a system that encompasses a wide range of requirements of healthcare system. To enhance... 

In this thesis, we first briefly review the basic concepts of stochastic processes. After reviewing and studying the dynamic equation that can explain a stochastic process, we show how one can find on a data-driven basis, the first-, second- and higher-order interactions between different subunits of a complex system by disentangling the dynamics of multivariate time series into stochastic and deterministic parts. Our data-driven approach is to detect different degrees of interactions obtained using conditional moments of Kramers-Moyal coefficients from unconditioned correlation functions and statistical moments of multivariate N-dimensional multivariate time series. Finally, we study the... 

Tipping points occur in diverse systems in various disciplines such as ecology, climate science, economy, sociology, and engineering. Critical thresholds in system parameters or state variables at which a tiny perturbation can lead to a qualitative change in the system exist in many subsystems in complex systems. These thresholds are called tipping points, and these subsystems are called tipping elements. Additionally, many systems with tipping points can be modeled as networks of coupled multistable subsystems. Domino-like tippings are called tipping cascades. Considering that these tipping cascades are primarily unprecedented, it is essential to study the dynamics and control of these... 

The aim of this thesis is to review the theory of deep learning with an experimental approach. In this thesis, we review researches that examine the impact of input selection on outputs in deep learning systems; Inputs we can control (samples, architecture, model size, optimizer, etc.) and outputs we can observe (the performance of the neural network, its test error, its parameters, etc.). Among the reviewed cases are the generalizability of deep learning systems, the effect of model components on its accuracy, interpolation and hyperparameters, as well as new phenomena in this field for which new frameworks have been defined  

One of the issues in the design and synthesis of the state or output feedback controller is the issue of fragility as the existence of disturbance in the coefficients of the designed controller causes instability in the controlled closed-loop system or leads to deterioration of its performance. The need to consider this point results in some novel approaches for designing the non-fragile controllers. Accordingly, defining criteria for measuring fragility is crucial.As well, due to the considered practical problems in the implementation of ordinary designed robust and optimal controllers, non-reachability of all the states of dynamical systems, and the boundedness for the quantity of the... 

This thesis presents an AI-powered framework for real-time energy and flexibility self- scheduling of a technical virtual power plant (TVPP). The TVPP integrates electric vehicle charging stations (EVCSs), energy storage systems (ESSs), photovoltaic (PV) panels, wind turbines (WTs), and distributed generators (DGs) to participate in whole- sale energy and flexibility markets while adhering to distribution network constraints. The framework employs a two-stage approach to maximize profitability. In the first stage, a deep learning model forecasts short-term flexibility requirements. In the sec- ond stage, optimization is divided into two interdependent sub-problems. The first sub- problem... 

Strongly Rayleigh measures are an important class of negatively dependent (repulsive) probability measures. These measures are defined via a geometric condition, called “real stability”, on their generating polynomials, and have interesting probabilistic properties. On one important property of negatively dependent measures is their rigid dependence structure. In other words, the it is impossible for these measures to have strong overall dependencies. In this thesis, we study two manifestations if this phenomenon: (1) paving property and (2) tail triviality. Informally, the paving property states that it is possible to partition the set of the components of every strongly Rayleigh random... 

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,... 

In this thesis, we give necessary and sufficient conditions for a simplicial com- plex with small codimension to satisfy the Serre’s condition (Sr) as well as the CMt property. We also give a connection between being (Sr) and being CMt. Also, we focus on the dimension of dual modules of local cohomology of Stanley– Reisner rings to obtain a new vector which contains important information on the Serre’s condition (Sr) and the CMt property as well as the depth of Stanley–Reisner rings. We prove some results in this regard including lower bounds for the depth of Stanley–Reisner rings. Further, we give a characterization of (d − 1)-dimensional simplicial complexes with codimension two which are... 

When we talk about the large-scale structure of the cosmos and the cosmic web, one of the important issues is to examine the location of the structures formed. We use statistical characteristics to check the cosmic web. Nearest neighbor distribution, two-point correlation, count in cell, etc. are some of the parameters that help us in this way. Our main problem in this project is to examine these statistical parameters in large scale structure for two different time periods. First, in the time period near the present age(z ≈ 0), we want to check these parameters for the dark matter halos that have formed the structure, and secondly, for the initial conditions and transition to high... 

In local transportation service companies, there is a need for estimation of arrival times (ETA) for customers. These companies not having the sufficient number of active online users, they are not able to obtain a proper estimation for ETA from that data. However, there are some overseas companies that have the established user base, and provide access to necessary data with a service fee. It is possible to solve this problem using the data which they provide.In this thesis, the problem of learning a hidden graph for the purpose of aiding local companies is explained and different aspects of it are introduced. In order to tackle this problem, first we define the idea of vertex separators,... 

In traditional networks, network management was performed locally at the data layer by switches and routers. This approach made it difficult to make optimal decisions with a comprehensive view of the network. Software-Defined Networks (SDNs), by separating the control layer from the data layer, introduced a centralized decision-making unit in the network. This decision-making layer, equipped with overall network information, makes decisions that enhance the overall efficiency of the network. Software-defined networks provide customizable traffic control by storing numerous rules in on-chip memories with minimal access latency. However, the current on-chip memory capacity falls short of meeting... 

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... 

Cells in their surrounding environment subjected to biochemical and biomechanical stimuli and respond to this signals. The mechanism by which cells transduce external mechanical signals into responses is called Mechanotransduction and cell use this mechanism to sense and respond to extracellular signals. Cells control processes like migration and division by mechanosensing. Cell migration is one of the most important responses to Cell-ECM interaction. This migration underlies many biological processes including embryogenesis, wound healing, tumorogenesis, morphogenesis and immune defense, so the study of cell migration is of particular importance. If there are external signals in cells... 

Few-shot learning methods have achieved notable performance in recent years. However, fewshot learning in large-scale settings with hundreds of classes is still challenging. In this dissertation, we tackle the problems of large-scale few-shot learning by taking advantage of pre-trained foundation models. We recast the original problem in two levels with different granularity. At the coarse-grained level, we introduce a novel object recognition approach with robustness to sub-population shifts. At the fine-grained level, generative experts are designed for few-shot learning, specialized for different superclasses. A Bayesian schema is considered to combine coarse-grained information with... 

Since the amount of data is increasing, it is important to reduce the size of components and circuits. For instance, bus routing problem, rectangle scape problem, and minimizing number of layers are important problems in printed circuit boards. Rectangle scape problem can be used as an estimation for minimizing number of layers problem. To minimize number of layers in this problem, we are given some axis-parallel rectangles inside a axis-parallel rectangular region. The objective is to extend one of the four boundaries of each rectangle in a certain direction such that all rectangles can be placed without any conflict in minimum number of layers. In this thesis, we analyze a common greedy... 

By Exploiting financial statements, ownership structure, and loan data for nonfinancial Iranian listed firms from 2007 to 2018, we investigate the impact of group structure on connected lending and its’ consequences on the firm investment. The group structure is formed, when a bank directly or indirectly owns one or multiple firms. Fist, our findings show that the vertical bank-firm connection, on average increases the probability and the amount of the loan supply by 2.53% and 78%, respectively. In return, the horizontal connection just increases the probability of the loan supply by 1.55%, and we find no evidence on its’ impact on the amount of the loan supply. Then, we investigate the...