Sharif Digital Repository

Traditional classification techniques of machine learning assume that data have stationary distributions. This assumption for recent challenges where tremendous amount of data are generated at unprecedented rates with evolving patterns, is not true anymore. Classification of data streams has become an important area of machine learning, as the number of applications facing these challenges increases. Examples of such data streams applications include text streams, surveillance video streams, credit card fraud detection, market basket analysis, information filtering, computer security, etc. An appropriate method for such problems should adapt to drifting concepts by revising and refining the... 

Isn’t our universe ,which includes everything that is causally connected to us, a tiny and negligible part of wider and diverse “multiple universes”? Although the multiverse idea has been considered by scientists recently, it had been discussed by philosophers and theologians in the past.
Many of our physical theories,whether being accepted like General Relativity, or speculative like string theory, lead naturally to the multiverse idea .The idea of multiverse is also brought forth to take care of the fine tuning observed in cosmology, in connection with the so – called anthropic principle.
This idea appeared in order to answer to some problems of quantum mechanics, string theory... 

Duplex stainless steels and advanced high strength steels are both known as new materials in automobile industry applications. Process of resistance spot welding is the main process in welding of auto-body structures. Hence investigating the welding metallurgy and weldability of similar and dissimilar joints of duplex stainless steels and advanced high strength steel are crucial. The purpose of this study is to analyse the relationship between welding process and geometric characteristic of weld nugget, phase transformations of Weld nugget and heat affected zone, mechanical properties (i.e force bearing capacity and energy absorption capability) and failure mode (interfatial or pullout). in... 

Mixed-criticality embedded systems, as the next-generation of safety-critical systems, are increasingly employed in industry due to consolidating functionalities with varying criticality levels onto the same computing platform. Also, cost reduction and performance improvement encourage adoption of multicore platforms in these systems. Technology scaling, battery-supplied design and heavy computation in mixed-criticality systems necessitates employing energy management techniques. Due to the degrading effects of these techniques on the system’s reliability, reducing energy consumption without sacrificing the reliability requirement is a vital challenge in mixed-criticality systems, in... 

Passive wideband reconnaissance receivers, listen to the radio and radar signals in the environment and estimate their important parameters such as direction of arrival (DOA). The wideband attribute for these receivers is to refer to their ability to receive wideband signals.Due to the use of processing gain, LPI radars reduce their transmitted signal power to see and not to be seen. On the other hand reconnaissance receivers can not have much processing gain, because they do not know the nature of the transmitted signal; Therefore, to increase the operational range of these systems, it is necessary to increase the signal to noise power ratio (SNR). This thesis examines the use of antenna... 

Ever increasing applications of embedded systems have motivated the designers to pay special attention to the design requirements of such systems. Among embedded applications, safety-critical systems have high reliability requirements as failures in such systems may endanger human life or result in catastrophic consequences. Embedded processors as the computation cores of embedded systems are very crucial from reliability point of view. This is because; a failure in the processor most probably leads to a system failure. One effective way to protect embedded processors against environmental faults is to use system level fault-tolerant techniques such as Master/Checker (M/C) or Triple Modular... 

In recent years, the use of embedded processors has grown increasingly in wide range of computer systems; so that most of manufactured CPUs are used in emebedded systems many of which are safty-critical systems such as medical devices, aircraft flight control, space systems, nuclear systems, etc. The incidence of failure in these systems can cause irreversible damages on human life, financial or environmental matters. Silicon process technology trends, such as reducing the threshold voltage, increasing the frequency and decreasing the size of transistors, not only caused increase in single-bit fault rate but also caused occurance of multi-bit faults. Due to importance ofcorrect operation of... 

Nowadays, one of the most important challenges in the design of digital circuits is their susceptibility to the strike of high energy particles which leads to the Single Event Transient (SET) and Multiple Event Transients (MET). In fact, technology scaling which results in lower supply voltage, higher operating frequency, and lower nodal capacitance, makes today’s digital circuits more susceptible not only to high energy particles but also to low energy particles. Moreover emerging deep sub-micron technologies and the integration of more cells in today’s chips have caused higher probability of MET occurrences. A lot of research has tried to reduce the soft error rate due to high energy... 

Wireless Sensor Networks (WSNs) are prone to faults due to battery depletion of nodes, where a node failure can disturb routing, as it plays a key role in transferring sensed data to the end users. A WSN may be partitioned into groups of nodes called clusters; for every cluster, there is a node, called Cluster Head (CH) that is responsible for collecting and transferring data from all nodes in that cluster. This means that the CH is a single point of failure; i.e., once a CH fails, the whole cluster will fail, which leads to inability of its members for transferring data outside the cluster. This thesis presents a Fault-Tolerant and Energy-Aware algorithm (FTEA), which prolongs the lifetime... 

Vulnerability of SRAM-based FPGAs to soft errors signals the importance of applying fault-tolerant methods in FPGAs used in safety-critical applications. Previous methods to protect SRAM-based FPGAs impose significant area and power overheads. Additionally, they suffer from susceptibility of configuration bits to Single Event-Multiple Upsets (SEMU). This thesis presents a Highly Available Fault-Tolerant Architecture (HAFTA) to protect SRAM-based FPGA designs against SEMUs in both configuration and user bits. In HAFTA, the entire design is duplicated and the main and replica flip-flops are compared at each clock cycle to detect any possible mismatch. To save the latest correct state of the... 

The increasing size of valuable data emphasizes the importance of applying reliability in storage systems. One way to protect storage system failures is using erasure codes. The advantages of using erasure codes are their low overheads and high reliability. Soft errors caused by high-energetic particles do not only corrupt data in the SSD-based storage systems, but also in the erasure codes. In this regards, it is important to protect erasure code implementations against soft errors. This thesis proposes a fault-tolerant implementation of erasure codes. The proposed method is based on the structure of each erasure code. This method is analytically evaluated on four erasure codes, i.e.... 

With continuous device down-scaling and increase in transistor counts on a chip, complementary metal-oxide-semiconductor (CMOS) technology has become extremely sensitive to soft errors. Soft errors are transient errors caused by energetic particles such as neutrons and alpha particles. An essential step to design a soft error tolerant digital system with minimal performance and power overheads is Soft Error Rate (SER) estimation of system components. Until recently, Single Event Upsets (SEUs) in latches and Filp-Flops (FFs) and Single Event Transients (SETs) in combinational logic parts of digital circuits were regarded as the main effects of particle strikes. However, with the emerging... 

Increasing complexity of embedded systems and the need for more computation powerhave directed designers toward using of multiprocessors. SRAM-based FPGAs are suitable platforms for implementation of multiprocessors due to thier low cost, fast time-to-market and re-configurability. FPGA-based multiprocessors are known as soft multiprocessors. The large area of SRAM-based FPGAs is occupied by configuration bits. Configuration bits are vulnerable to high energy particles that can lead to soft errors. In this regards, it is of decisive importance to protect soft multiprocessors against soft errors. This thesis proposes a fault-tolerant method for soft multiprocessors that can detect and... 

The use of flash-based Solid State Disks (SSDs) has been increased in the past decades because of their high performance and low power consumption. As SSDs have no moving mechanical parts, they are more reliable than HDDs. SSDs have specific errors, such as read, program, and erase disturbs, data retention, and endurance. It is so common in storage systems to apply an array of disks called Redundant Array of Independent Disks (RAID) together for the aim of reliability and performance. Erasure codes are one of the main methods applied in storage systems for the aim of reliability. Erasure codes increase the reliability of storage systems by protecting them against disk failures and sector... 

One of the major challenges that threat the reliability of NoC-based systems is Crosstalk fault. Some effects of crosstalk fault such as Rising/falling delays and rising/falling speed-ups lead to the variations in channel delay, incorrect data transmission, and extra power consumption in NoC communication channels. Crosstalk fault is data dependent and the intensity of this fault seriously depends on the transition patterns appearing on the wire during the data traversal between processing elements. Most of mechanisms tackling crosstalk fault that are discussed in literature are based on 3-wire delay model. In 3-wire delay model, one wire is considered as victim wire and classification of... 

Wireless Sensor Networks (WSNs) consist of sensor nodes that are connected to each other and are widely used in many applications to acquire and process information. WSN nodes are battery powered, therefore energy management is a key factor for long live network. Node radio transceiver unit uses the most part of energy resource of the node and as a result limits the network lifetime. One way to prolong lifetime of network is to utilize routing protocols to manage energy consumption. To have an energy efficient protocol, we can apply cluster organization on the network, where sensor nodes are partitioned into groups called cluster. Then, the whole cluster data is sent through Cluster Head... 

Three-dimensional network-on-chip (3D NoC), the combination of NoC and die-stacking 3D IC technology, is motivated to achieve lower latency, lower power consumption, and higher network bandwidth. 3D NoCs reduce length of communication link and transmition delay by die stacking. However, the length of heat conduction path and power density per unit area increase as more dies are stacked vertically. High temperature results in longer propagation delay and increases the leakage power. If heat production in the chip is not managed, low cooling rate for these layers is a threat for the chip performance and reliability. To solve the thermal problems, thermal management schemes were proposed in the...