Sharif Digital Repository

By the improvement of machine learning methods specially the Deep Learning in the last decade, there were expanding usage of these methods in Language Modeling task. As the essence of a language model is more basic, recently huge networks are trained with language model objective but fine-tuned on target tasks such as Question Answering, Sentiment Analysis and etc. which is a promising sign of its importance and usage in even other NLP tasks. However, this task still has severe problems. The Teacher Forcing based methods, suffer from the so-called exposure bias problem which is due to the train/test procedure discrepancy. Some solutions such as using Reinforcement Learning which has high... 

Most elderly and physically disabled people suffer from lack of functionality and dexterity in their elbow or wrist. These disabilities are due to the damages mostly caused by sport surgery, spinal surgery, or stroke. Therefore, design of an assistive exoskeleton robot for upper limb movements seems necessary. The purpose of this study is to design, fabricate and deliver a control algorithm for an assistive wearable robot. The robot has five degrees of freedom in order to help the flexion/extension and abduction/adduction shoulder. Dynamic and kinematic model of elbow, forearm, and wrist is developed to determine the amounts of torques which are required in the joint actuators mounted on the... 

In this project, a ceramic biocompatible coating was applied on Titanium which is included Hydroxyapatite by AC Plasma Electrolyte Oxidation method. First, coating was performed in 5 different solutions and various duration in 500mA/Cm2 as a current density. Then, by microscopic and macroscopic images and size of porosity and also thickness of coat, 3 solutions were chosen for adding nanoparticles: Solution 1 which includes NaH2PO4 and Ca(CH3COO)2 , Solution 3 which includes Ca(CH3COO)2 and Na-Beta G and Solution 4 which includes Ca(H2PO4)2, HMP, NA2(EDTA) and Ca(CH3COO)2 . Also the time 10 minutes was chose as a appropriate time. The average of the porosities size in solution 1 was about... 

Solving the visual symbol grounding problem has long been a goal of modern aritificial intelligence. Due to recent breakthroughs in deep learning methods for natural language processing and visual interpretation tasks‚ the field now seems to be as near to achieving this goal as it ever was. Also recent progress in using recurrent neural netowrks (RNNs) for image description‚ has motivated the exploration of their application for video description tasks. However, while images remain static‚ interpreting videos require modeling complex dynamic temporal sturctures and then properly integrating that information into a natural language description. Recurrent neural networks can be both used to... 

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... 

Despite the remarkable success of deep learning algorithms in recent years, it still has a long way to reach the status of human natural intelligence and to acquire the expected self-autonomy. As a result, many researchers in this field have focused on the development of these algorithms while taking inspiration from human cognitive behaviors. One of the disadvantages of current algorithms is the lack of their ability to learn in a continual manner while deployed in the environment. More precisely, deep learning models are not able to gradually gather knowledge from the environment and if they are in a situation of limited access to data, they will suffer from catastrophic forgetting; a... 

Recent advances in deep neural networks have shown significant potential; however, they still face challenges when it comes to non-stationary environments. Continual learning is related to deep neural networks with limited capacity that should perform well on a sequence of tasks. On the other hand, studies have shown that neural networks are sensitive to covariate shifts. But in many cases, the distribution of data varies with time. Domain Adaptation tries to improve the performance of a model on an unlabeled target domain by using the knowledge of other related labeled data coming from a different distribution. Many studies on domain adaptation have optimistic assumptions that are not... 

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... 

Nowadays, one of the challenges in social robotics is to navigate the robot in social environments with moving elements such as humans. The purpose of this study is to navigate the Arash 2 social robot in a dynamic environment autonomously without encountering moving obstacles (humans). The Arash 2 robot was first simulated in the Gazebo simulator environment in this research. The simultaneous location and mapping (SLAM) technique was implemented on the robot using a lidar sensor to obtain an environment map. Then, using the deep reinforcement learning approach, the neural network developed in the simulation environment was trained and implemented on the robot in the real environment. The... 

Statistical studies indicate the fact that in Iran, like many other developing countries, a significant percentage of buildings are made of traditional materials and most of them are adobe buildings especially in rural areas. Furthermore, geological data indicates that Iran is located in alps-Himalayas seismic zone and exposed by destructive earthquakes and adobe buildings suffer the most damages comparing with other type of the structures. According to these explanations, this research deals with adobe buildings behavior and provides solutions for reinforcing them and improving the seismic response of these buildings. These studies include numerical modeling with finite elements methods by... 

Ensuring the reliability of Large Language Models (LLMs) in complex reasoning tasks remains a formidable challenge, particularly in scenarios that demand precise mathematical calculations and knowledge-intensive open-domain generation. In this work, we introduce an uncertainty-aware framework designed to enhance the accuracy of LLM responses by systematically incorporating model confidence at critical deci- sion points. We propose an approach that encourages multi-step reasoning in LLMs and quantify the confidence of intermediate answers such as numerical results in math- ematical reasoning and proper nouns in open-domain generation. Then, the overall confidence of each reasoning chain is... 

In this thesis we consider the problem of distributed computation by many processors.We mainly concentrate on matrix multiplication problem in this thesis because of its importance. A distributed system consists of N worker processors and one master processor. The master processor should distribute the computation between workers and after computation in each of them, collect the results. In this thesis, we are going to mitigate the effect of straggler processors by using coding methods. Straggler processors can cause delays in the computation time.In this thesis, we firstly introduce a method to multiply any number of matrices in each other. The proposed method occurred in one shot without... 

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.... 

The main goal of current investigation was partial reduction of Fe2O3 available in Hematite-rich ore to Magnetite by Fludization Bed method. The reductive gas, CO, was produced by incomplete combustion of Acetylene and Airmixed in inappropriate ratio. X-Ray Diffraction and X-Ray Florescence analyses were implemented in order to conducting qualitative analysis of present phases before and after reduction of the ore. Quantitative analysis of the same phases was also conducted by Titration. The latter analysis revealed the precise magnitude of available iron in each of the oxide containing phases. The chemical composition of the reactor gas was also determined before and after reduction... 

Abstract The problem of competetive facility location can be defined as follows: There are a number of customers in the form of points in space, and two players arrange a number of facilities in the form of points in space, given some limitations, respectively. Each customer’s connection to each facility has a cost for the customer and an advantage for the facility, and each customer wants to be connected to at most one of the facilities which has the lowest cost for him. The goal is to find the strategy of placing the facilities and determining the cost which the facility receives from the customer, in such a way that the player’s profit is maximised.In this thesis, we first sought to... 

Zero knowledge proofs allow a person (prover) to convince another person (verifier) that he has performed a specific computation on a secret data correctly, and has obtained a true answer, without having to disclose the secret data. QAP (Quadratic Arithmetic Program) based zkSNARKs (zero knowledge Succinct Non-interactive Argument of Knowledge) are a type of zero knowledge proof. They have several properties that make them attractive in practice, e.g. verifier's work is very easy. So they are used in many areas such as Blockchain and cloud computing. But yet prover's work in QAP based zkSNARKs is heavy, therefore, it may not be possible for a prover with limited processing resource to run... 

Lightweight clients are a type of blockchain users who do not store all the blocks in the blockchain due to limited resources. These users store only a small part of each block and when needed, request transactions from full nodes that store the entire blockchain. These users have no role in block validation and only want to receive transactions related to their addresses with proof of the inclusion in the block from full nodes.Since light clients rely on full nodes for receiving transactions, their privacy against full nodes is important. The current implementation of Bitcoin uses Bloom filters for privacy, but this offers very little privacy to the users.In this thesis, we study the...