Loading...
				
	
				
	
								
								
			Search for:
			
					    			jasemi--masoumeh-sadat
					    	
		
							
	    		0.121 seconds
			
			
				Total 112 records
			
		Improving the Performance of Non-Volatile Memory based CNN Accelerators
, Ph.D. Dissertation Sharif University of Technology ; Hessabi, Shaahin (Supervisor) ; Bagherzadeh, Nader (Co-Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Today convolutional neural networks (CNN) are very popular due to their high accuracy and robustness. As the size and complexity of CNNs grow, the demand for larger on-chip memories also increases. Given that off-chip access memory has a high cost, one solution is to enlarge on-chip caches by employing emerging multi-level Cell (MLC) STT-RAMs. This memory provides higher capacity at the cost of lower reliability. The root cause of low reliability of MLC STT-RAM is failure of read and write operations. However, MLC STT-RAM and CNNs are perfect matches in the sense that in one hand MLC STT-RAM provides higher capacity, on the other hand CNN can tolerates moderate level of inaccuracy and low... 
				
					On competence of vendor managed inventory in supply chains using basic mathematical inventory models
, Article Scientia Iranica ; Vol. 21, issue. 3 , 2014 , p. 1061-1071 ; Haji, A ; Gharibi, M ; Sharif University of Technology
								
								
					2014
				
							
				
		
							Abstract
				
					
			
		
										
				In this study, a two-echelon single-vendor supply chain is selected to do a cost-based comparison between short-term performances of a Vendor Managed Inventory (VMI) and a Retailer Managed Inventory (RMI). While the inventory costs include ordering and storing expenses, the rate of consumption and the price of goods are constant, the rate of production and pace of transportation are infinite and shortage is not allowed. This paper, after a comprehensive literature review, is followed by three cases of single retailer, n-retailer and two-retailer chains. Unlike the second case, in the first case, VMI shows an absolute superiority to RMI, and this is the reason for devising the third case in... 
				
				
				
					Reliable and energy efficient MLC STT-RAM buffer for CNN accelerators
, Article Computers and Electrical Engineering ; Volume 86 , 2020 ; Hessabi, S ; Bagherzadeh, N ; Sharif University of Technology
								
					Elsevier Ltd 
				
								
								
					2020
				
							
				
		
							Abstract
				
					
			
		
										
				We propose a lightweight scheme where the formation of a data block is changed in such a way that it can tolerate soft errors significantly better than the baseline. The key insight behind our work is that CNN weights are normalized between -1 and 1 after each convolutional layer, and this leaves one bit unused in half-precision floating-point representation. By taking advantage of the unused bit, we create a backup for the most significant bit to protect it against the soft errors. Also, considering the fact that in MLC STT-RAMs the cost of memory operations (read and write), and reliability of a cell are content-dependent (some patterns take larger current and longer time, while they are... 
				
				
				
					Enhancing reliability of emerging memory technology for machine learning accelerators
, Article IEEE Transactions on Emerging Topics in Computing ; Volume 9, Issue 4 , April , 2021 , Pages 2234-2240 ; 21686750 (ISSN) ; Hessabi, S ; Bagherzadeh, N ; Sharif University of Technology
								
					IEEE Computer Society 
				
								
								
					2021
				
							
				
		
							Abstract
				
					
			
		
										
				An efficient and reliable Multi-Level Cell (MLC) Spin-Transfer Torque Random Access Memory (STT-RAM) is proposed based on a Drop-And-Rearrange Approach, called DARA. Since CNN models are rather robust, less important bits are dropped, allowing important bits to be written in safe and reliable Single-Level Cell mode. Also, bits are rearranged to make the representation better aligned with memory cell characteristics. Bits with higher impact on the features value are stored in safer bit positions reducing the chance of read/write circuits to malfunction. Experimental results show that our approach provides comparable to error-free scenario reliability level, while doubling the bandwidth and... 
				
				
				
					Partition pruning: Parallelization-aware pruning for dense neural networks
, Article 28th Euromicro International Conference on Parallel, Distributed and Network-Based Processing, PDP 2020, 11 March 2020 through 13 March 2020 ; 2020 , Pages 307-311 ; Albaqsami, A ; Jasemi, M ; Bagherzadeh, N ; Sharif University of Technology
								
					Institute of Electrical and Electronics Engineers Inc 
				
								
								
					2020
				
							
				
		
							Abstract
				
					
			
		
										
				As recent neural networks are being improved to be more accurate, their model's size is exponentially growing. Thus, a huge number of parameters requires to be loaded and stored from/in memory hierarchy and computed in processors to perform training or inference phase of neural network processing. Increasing the number of parameters causes a big challenge for real-time deployment since the memory bandwidth improvement's trend cannot keep up with models' complexity growing trend. Although some operations in neural networks processing are computational intensive such as convolutional layer computing, computing dense layers face with memory bandwidth bottleneck. To address the issue, the paper... 
				
				
				
					Cluster-based adaptive SVM: a latent subdomains discovery method for domain adaptation problems
, Article Computer Vision and Image Understanding ; Volume 162 , 2017 , Pages 116-134 ; 10773142 (ISSN) ; Jamzad, M ; Sharif University of Technology
								
								
					2017
				
							
				
		
							Abstract
				
					
			
		
										
				Machine learning algorithms often suffer from good generalization in testing domains especially when the training (source) and test (target) domains do not have similar distributions. To address this problem, several domain adaptation techniques have been proposed to improve the performance of the learning algorithms when they face accuracy degradation caused by the domain shift problem. In this paper, we focus on the non-homogeneous distributed target domains and propose a new latent subdomain discovery model to divide the target domain into subdomains while adapting them. It is expected that applying adaptation on subdomains increase the rate of detection in comparing with the situation... 
				
				
				
					Indoor Noise Pollution Modeling: A Case Study on Resalat Tunnel
, M.Sc. Thesis Sharif University of Technology ; Vaziri, Manouchehr (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Noise or noise pollution, is defined as an unwanted sound, produced mostly by motor vehicles in urban areas. If the noise exceeds its limit, adverse effects on users and citizens are expected. Thus the assessment of the noise level and its comparison with allowable values is mandatory. Traffic induced noise depends upon different factors such as traffic, weather, and, sound barriers. Noise pollution study has a long history all over the world. Many studies, mostly representing the outdoor noise pollution models, have been done in recent years. This paper presents an assessment of traffic-induced noise in a heavily traveled tunnel, locally referred to as Resalat tunnel, which is located in... 
				
					The Effect of Travel Time Reliability on Morning Peak Travellers route choice in Large Metropolitan Area
, M.Sc. Thesis Sharif University of Technology ; Poorzahedy, Hossein (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Data collected from street networks show that travel time in urban street networks is a stochastic variable, whose range of variation increases with the congestion level. Work and school trips comprise the major part  of morning peak volumes, whose common characteristic is that they need to be made on time. For this reason, these travelers are pessimistic in their route choice, and try to choose routes which are more reliable. Travel time standard deviation is one of the common measures for estimating travel time reliability that is related to properties of day-to-day travel time distribution on a particular route. Based on the empirical data collected from Tehran street network, this paper... 
				
					Design and Implementation of a FlexRay Transceiver
, M.Sc. Thesis Sharif University of Technology ; Atarodi, Mojtaba (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				With the advancement of automotive industry, the number of electronic systems that need to communicate with each other has increased. An effective strategy to meet these growing demands is using bus-based protocols, which can establish safe communication among numerous nodes with reasonable cost, weight, and space requirements. The objective of this thesis is the design and implementation of a transceiver based on the FlexRay protocol. The significance of this protocol lies in its ability to provide higher data transmission rates compared to other protocols like CAN and LIN, and its role as the foundational communication medium among various domains in modern vehicle architectures. In this... 
				
					Using Information Beyond Text to Generate Language Embedding Vectors
, M.Sc. Thesis Sharif University of Technology ; Sameti, Hossein (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				In this thesis, we introduce a novel Artificial Intelligence (AI) system inspired by the philosophical and psychoanalytical concept of imagination as a ``Re-construction of Experiences". Our AI system is equipped with an imagination-inspired module that bridges the gap between textual inputs and other modalities, enriching the derived information based on previously learned experiences. A unique feature of our system is its ability to formulate independent perceptions of inputs. This leads to unique interpretations of a concept that may differ from human interpretations but are equally valid, a phenomenon we term as ``Interpretable Misunderstanding". We employ large-scale models,... 
				
					Entanglement Distillation for Weakly Entangled States
, M.Sc. Thesis Sharif University of Technology ; Mirkamali, Maryam Sadat (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Quantum entanglement is an essential resource in quantum information processing, such as quantum communication and quantum computing. Although entanglement is critical for practical implementations in these fields, creating and transmitting entanglement without altering the states is challenging. External noise may destroy or weaken the entanglement. Consequently, there is a need for methods to improve entanglement. Entanglement distillation is an effort to address this issue. It is a process where, probabilistically, a smaller number of strongly entangled states are created from a fixed number of noisy entangled states. This process is carried out using only local operations and classical... 
				
					Differentially Detection in UWB Communicatio Systems without Reference Transmission
, M.Sc. Thesis Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				In recent years, UWB multiple access communication systems have received a great attention due to its many advantages. The most important advantages of a UWB system can be summarized as its high data rates, low power consumption, fading immunity and high precision routing. One of the important problems in a UWB system implementation is in its receiver part in which in order to use a coherent RAKE receiver, the channel coefficients have to be estimated. Since UWB channel is a dense multipath fading channel with a very high number of resolvable paths, the channel coefficient estimation based on using training sequence is very complicated. To resolve this problem, transmit reference (TR)... 
				
					Cooperative Communications in Free-Space Optical Networks
, Ph.D. Dissertation Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				With recent successes in laboratory tests, and in atmosphere and space demonstrations, there is no doubt that the free-space optical (FSO) communications is ready for operational deployment. Large bandwidth, unlicensed spectrum, excellent security, and quick and inexpensive setup are among its most attractive features. Despite their attractive features, FSO communications suffer from several challenges in practical deployment; the major of them is fading or scintillation and path loss. To overcome such limitations, Channel Coding along with MIMO and Cooperative techniques have been proposed. Although the promising effects of cooperative transmissions in RF communications have greatly been... 
				
					Harnessing Interference in Cooperative Communication Networks
, Ph.D. Dissertation Sharif University of Technology ; Nasiri-Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				The compute-and-forward (CMF) method has shown a great promise as an innovative approach to exploit interference towards achieving higher network throughput. Here, instead of recovering single messages, the relays attempt to reliably decode (compute) and pass an integer linear combination of the transmitted messages, referred to as an equation, to the destination. By receiving sufficient equations,the destination can solve the linear equation system to recover the desired messages.Two main challenges of the CMF methods are the computational complexity and the rank failure problem at the destination. In this thesis, to decrease the implementation as well as the computational complexity of the... 
				
					Multiple-Access Techniques for Wireless Indooor Infrared Communication with Dispersive Link
, Ph.D. Dissertation Sharif University of Technology ; Nasiri-Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Infrared wireless local area networks are a suitable choice for a fast and cheap wireless network, providing rates over 10 Mbps with access to unregulated bandwidth. Among multiple-access techniques for infrared wireless networks, CDMA1 has attracted the most interest, due to its simple structure and power ecency. However, spreading the spectrum in CDMA prohibits exploiting the low-rate strong channel codes. The other limiting factor in infrared networks is the channel dispersion. In this thesis, these two issues are considered and an internally coded scheme for time-hopping CDMA, suitable for wireless infrared communication is proposed and with accurate modelling of infrared dispersive... 
				
					Spectrum Sensing in Cognitive Radio Networks
, Ph.D. Dissertation Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				In this thesis, we consider the problem of spectrum sensing in cognitive radio networks. First, the collaborative energy detectors based spectrum sensing are investigated in the case of known noise variance for two models of primary user (PU)  signal, i.e. random and unknown deterministic signals. Since the derived optimum collaborative energy detector requires the signal-to-noise ratio (SNR) of secondary users (SU) and it has complex structure, the generalized likelihood ratio (GLR) detector is proposed for both models of PU signal which leads to the same decision rules for both models. Simulation results show that the performance of the proposed GLR detector is near to that of optimal... 
				
					Efficient Transmission and Reception in Wireless Communication Systems based on NOMA
, Ph.D. Dissertation Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Multiple access technologies are one of the key features of different generations of wireless communication. From the first generation to the fourth generation, orthogonal multiple access (OMA) schemes have been utilized to serve multiple users in different orthogonal resource blocks in time, frequency or code domain. However, the new application scenarios for the fifth generation impose the requirements of higher system throughput, massive connectivity, and low latency, which can not be fulfilled with OMA schemes. Therefore, non-orthogonal multiple access (NOMA) schemes have attracted increasing attention in both industry and academia. In NOMA, multiple users may occupy the same resource... 
				
					Analysis of Cache-aided Wireless Communications Systems in the Presence of IRS
, M.Sc. Thesis Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Future mobile networks are expected to not only bring more wireless connectivity and higher data  rates, but also to support several high-quality services including low-latency, ultra-reliable, and energy-efficient communications.  To support these needs, intelligent reflecting surfaces (IRSs) have recently attracted a lot of attention due to their potential for reconfiguring the wireless channel and are considered as a promising solution for future mobile generations. On the other hand, the exponential growth of network traffic, and its non-uniform distribution result in network congestion during peak-traffic times and network underutilization during off-peak times. Caching or content... 
				
					Deformation and Annealing Behavior of an Austenitic Stainless Steel
, M.Sc. Thesis Sharif University of Technology ; Serajzadeh, Siamak (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				In this research, non-isothermal annealing process at temperatures ranging between 400°C to 830 °C was investigated in warm and cold-rolled austenitic stainless steel 304L. First, the wide variety of austenitic to martensitic structure was produced by the multi-pass rolling process at room temperature  -8 °C, 200 °C, and 280°C. The n exponent in Olson-Cohen equation was governed 2.83 after non-linear curve fitting to assess the fraction of deformation-induced martensite as a function of strain. Afterward, Abaqus/Explicit method was utilized to achieve the stored energy and plastic strains distribution through the thickness. Then, Cellular automata and Galerkin-finite element models were... 
				
					Interference Avoidance Algorithms in Cognitive Radio Networks and Performance Analysis
, M.Sc. Thesis Sharif University of Technology ; Nasiri Kenari, Masoumeh (Supervisor)
							Abstract
				
					
		
		
		
		
		
		
										
				Cognitive radio faces specific challenges due to its special spectrum access scheme. Providing required QoS and preventing irrecoverable interference on the primary users is among the crucial requirements. Beam forming is one of the interference avoidance schemes utilized in MIMO networks. Most of the studies on this scheme assume that each symbol is transmitted via vectors in order to grant desired beam forming and specific combinations in the receiver. Since transmitting multiple symbols at every transmission can provide higher data rates, it seems necessary to exploit beam forming through matrixes instead of vectors. In this thesis, by neglecting the interference of the secondary user,... 
				
					