Abstract

The human brain is a highly complex network of interconnected neurons exhibiting emergent phenomena. The theory of the brain functioning at a critical state - near a second-order phase transition where its capabilities for memory, learning, and information processing - has gained widespread popularity in recent years. This criticality has often been likened to the critical state observed in the Ising model, which is influenced by a temperature parameter. Neural dysfunction due to neurodegenerative conditions such as Alzheimer’s Disease influences the proximity to this critical state. We study the disrupted proximity to criticality of the dysfunctional Alzheimer’s brain using a spin-lattice pairwise maximum entropy model where connectivity is inferred from resting-state fMRI data. The closeness to criticality is quantified through the critical temperature, which is derived by analysing the susceptibility of the system. We find an insignificant difference between the two classes, which we attribute to the dependence of susceptibility on spatial dynamics, with the hypothesis that the existence of a spin-glass-like criticality could hinder the sensitivity of this metric. To characterise the nature of brain criticality, we explore the spatial characteristics and temporal correlations of criticality for both normal and Alzheimer’s brains. We find a weak distinction in the spatial dynamics between the two classes - the sizes and number of spatial domains formed, as well as the self-averaging behaviour of the system, were not found to vary significantly. However, we notice a stretched exponential form in the temporal relaxation times, which is characteristic of spin-glasses. We find a significant difference between the relaxation behaviour of Alzheimer’s and Cognitively Normal subjects. In spin-glass systems, a third phase exists near the phase transition boundary - the spin-glass phase. This phase is characterised by frustration and competing interactions. It allows for the emergence of complex properties such as long-term memory, which are features not found in simple Ising-like criticality. Despite the difference in closeness to criticality, the lack of a significant distinction in the spatial correlation lengths can be attributed to local frustrations, which might not allow the system to reach equilibrium domain distributions. Based on the differences observed between the spatial and temporal correlations, we conclude that the criticality found in the brain is more akin to a spin-glass criticality, which is more sensitive to temporal rather than spatial measures.

Abstract

Education plays a vital role in shaping up one’s life. Education in early childhood includes learning from different activities where counting and other concepts of mathematics act as major building blocks. Mathematics is not just a subject to be taught in schools, but also it has myriad applications in our daily lives such as counting the total number of stationery items, calculating their individual prices, and adding them up for a purchase made in a grocery shop. This kind of real world situations are posited in mathematical word problems which are an essential part of a child’s learning that requires natural language understanding (NLU) as well as knowledge of mathematical operations. Mathematical Word Problem Solvers can assist both students and teachers. It is a challenging field and this thesis attempts to provide NLP solutions for math word problems in English and Indian languages. Our approach for developing word problem solvers follows a pipeline. For a word problem, first, we identify the relevant operands and required operations. In the second step, we form an equation from these identified components. The first two stages can be combined to generate equations at once using neural network based approaches. At the last stage, the equation is solved by a mathematical solver to get the final solution. We focus primarily on the first two stages of this pipeline. We developed solvers using three kinds of approaches: frame based, composition of classifiers based, and end-to-end neural based. We also shed light on the limitations of the current automatic solvers with respect to the data. We designed different data augmentation techniques to overcome the data scarcity problem. As a part of resource building, we developed word problem datasets and solvers for Indian Languages. In addition to this, we compared different models related to our developed approaches. We empirically show the difference in generation of various equation notation types. For this study, we present the results of equations in infix, postfix, and prefix notations. We also show two natural language processing tasks where components of word problem solvers can be utilized. In the first task, we studied the impact of simple number based pre-processing on the performance of machine translation systems. In the second task, we analyze speech transcript texts to extract equation spans. This kind of text is present mainly in transcripts from mathematical domain. For achieving this, we develop an equation identifier and convertor involving math notations for transcripts. This can make the transcripts easily readable for transcripts which have wide usage of mathematical terms.

Abstract

Recent advancements in the field of advanced driver assistance systems (ADAS) have led to increased demands for mmWave (millimeter-wave) radars operating within the 24 GHz and 77 GHz frequency bands. Among the various radar techniques, the Frequency Modulated Continuous Wave (FMCW) method has gained considerable popularity among mmWave radars. This approach relies on utilizing frequency synthesizers with low phase noise and high bandwidth to attain higher precision and accuracy in radar applications. Of particular interest in recent years is the growing demand for FMCW radars that operate within the 76-81 GHz frequency band due to their exceptional precision and accuracy, thanks to their wide bandwidth. These radars are well-suited for a range of applications, including autopilot systems and ADAS. The primary focus of this thesis is to present research efforts aimed at enhancing the performance of FMCW radars at the circuit level, with a strong emphasis on the design, optimization, and simulation of Voltage-Controlled Oscillators (VCOs). VCOs play a critical role in radar systems, and two crucial design considerations are minimizing phase noise and reducing power consumption. These parameters directly impact the sensitivity and energy efficiency of the radar system. Various VCO topologies are explored, primarily based on LC oscillators, and comprehensive optimization endeavors are done to simultaneously minimize phase noise and power consumption. The VCO designs are implemented using TSMC 65 nm CMOS technology, allowing us to investigate innovative techniques for mitigating phase noise in VCOs. In summary, this thesis encompasses recent developments in FMCW radar technology, specifically addressing improvements at the circuit level by focusing on the design and optimization of VCOs.

Abstract

Damage caused by an earthquake depends not just on the intensity of an earthquake but on the engineering and construction practices of the region. Past earthquakes in Asian countries have highlighted inadequate construction practices and caused huge property and life loss, indicating the severe need to retrofit existing structures. For this, an understanding and quantification of damage are required to assess the amount of damage. Many researchers have proposed many indices; out of those, energy-based indices are chosen for this study. In literature, energy-based damage indices have progressed well but with a gap in mapping of local-to-global indices though they can be computed at local and global levels. This study bridges the gap between global and local energy-based indices using weighting factors, which further proposes sequence of strengthening. Strengthening activities shall be done first at the higher weighted members to increase the capacity of the building immediately and then to the other members. The proposed strengthening sequence is proposed to make it more economical and efficient. This study is conducted on existing buildings, mainly on gravity load-designed buildings (Type G) and precode buildings of IS 1893 designed as per the 2002 version (Type P). These buildings represent a significant stock of RCC MRF buildings on site that should be strengthened. The members contributing significant damage are identified in existing buildings, different weights are allotted to members as per their contribution to global energy dissipation. Nonlinear static analysis is conducted to quantify the damage. The methodology adopted to compute the contribution of local damage to the exterior and interior columns is verified with nonlinear time history analysis with eleven earthquakes. In gravity, load-designed buildings, higher weight for exterior columns is observed in the energy dissipation capacity of the building than interior columns. However, in precode buildings of IS1893, the significance of exterior and interior columns is similar. Storey-wise weights are computed using energy dissipated by each hinge in that storey. Therefore, the highest weighted stories are identified for each type of building. Damage is distributed within stories as a parabolic curve. In type G buildings, as the height of the building increases, parabolic distribution changes from convex to concave, and the maxima location of the parabola shifts from bottom to middle stories. In type P buildings, parabolic damage distribution remains convex or like a straight line. As the height of the building increases, damage shifts to upper stories in a convex parabolic shape. An increase in the storey height of a building does not change the damage distribution pattern and the quantity of damage. The sequence for strengthening activities is proposed as per the computed weighting factors. Strengthening activities shall be done first at the higher-weighted members to increase the capacity of the building immediately. Further, a strengthening sequence is proposed as per the weighting factor in descending order for regular RCC buildings. Therefore, proposals made in the study would increase the efficacy of strengthening activities.

Abstract

Electrical load forecasting is a prime element for power system planning, scheduling, economic dispatch, unit commitment, maintenance, and electricity market. In practice, the electrical load is having nonlinear behavior and electrical load demand depends upon various factors like the type of day (day of the week etc.), weather variables like temperature, humidity, and the load on the previous day etc. Electrical load forecasting can become more accurate by considering these dominant factors. Soft computing techniques have been successfully incorporated into a lot of scientific and engineering problems during recent years. Soft computing methods are capable of dealing with nonlinearity and these techniques are suitable candidates for electrical load forecasting problems. In this thesis work, soft computing techniques like fuzzy logic and bio-inspired computation techniques are chosen for similar day based electrical short term load forecasting. The fuzzy logic is able to model uncertain and ambiguous data. The fuzzy logic system gives better forms of rule expressions, reasoning logic like human thought. This work presents fuzzy logic based Short Term Load Forecasting (STLF) by selecting similar days, which are found from previous data by using Mahalanobis distance norm. The fuzzy logic is chosen in the correction of similar days of the forecast day. The average of the corrected hourly loads of the similar days achieved through the Mahalanobis distance norm of the forecast day is then considered as the hourly load of the forecast day. The parameter limits of fuzzy membership functions are determined and optimized through different optimization techniques like Heuristic Approach, Particle Swarm Optimization (PSO), New Particle Swarm Optimization (NPSO) and Evolutionary Particle Swarm Optimization (EPSO). In the first approach, Fuzzy membership parameter values are tuned in a heuristic way. Another approach for the optimization of fuzzy parameter values is to use PSO. It simulates the bird flocking behavior where each particle fitness value can be evaluated in a search space by using the fitness function. The particle’s direction of flying can be decided by its velocity and position. PSO function updates the latest particle position by using an optimization equation based on the global best of the previous iteration provided that it is better than the previous one. NPSO technique is also attempted for the optimization of fuzzy parameter values. Each particle tries to leave its previous worst position and the previous worst position of its group. NPSO function modifies the latest particle position using the optimization equations based on the global worst of the previous iteration provided that it is better than the previous one. The inclusion of the evolution process in PSO is providing the Evolutionary Particle Swarm Optimization (EPSO). EPSO has the properties of Replication, Mutation, Reproduction, Evaluation and Selection. In this thesis, EPSO is also used for the optimization of fuzzy parameter values. These optimized fuzzy parameter values are incorporating in Fuzzy Inference System (FIS) and used for forecasting the load of forecast month. It is difficult to forecast a load of anomalous days like holidays, religious days and special days because their load profile is different when compared to normal days. Load forecasting for anomalous days is also a challenging job because of small number of data availability compared to the availability of data for normal weekdays and weekends. Hence, various methodologies are developed for anomalous day load forecasting and presented in the present research work. The short term load forecasting algorithms are developed for each optimization technique and also forecasting methodologies for various case studies of anomalous day load forecasting and these are tested by doing simulation studies on real-time data and results obtained are found good, which have been presented in this thesis.

Abstract

The exponential performance growth guaranteed by Moore’s law has started to taper in recent years. At the same time, emerging applications like image processing demand heavy computational performance. These factors inevitably lead to the emergence of domain-specific accelerators (DSA) to fill the performance void left by conventional architectures. FPGAs are rapidly evolving towards becoming an alternative to custom ASICs for designing DSAs because of their low power consumption and a higher degree of parallelism. DSA design on FPGAs requires careful calibration of the FPGA compute and memory resources towards achieving optimal throughput. Hardware Descriptive Languages (HDL) like Verilog have been traditionally used to design FPGA hardware. HDLs are not geared towards any domain, and the user has to put in much effort to describe the hardware at the register transfer level. Domain Specific Languages (DSLs) and compilers have been recently used to weave together handwritten HDLs templates targeting a specific domain. Recent efforts have designed DSAs with image-processing DSLs targeting FPGAs. Image computations in the DSL are lowered to pre-existing templates or lower-level languages like HLS-C. This approach requires expensive FPGA re-flashing for every new workload. In contrast to this fixed-function hardware approach, overlays are gaining traction. Overlays are DSAs resembling a processor, which is synthesized and flashed on the FPGA once but is flexible enough to process a broad class of computations through soft reconfiguration. Image processing algorithms vary in size and shape, ranging from simple blurring operations to complex pyramid systems. The primary challenge in designing an image-processing overlay is maintaining flexibility in mapping different algorithms. This thesis proposes a domain-specific language (DSL)-based compiler, Flowpix, for image processing. Flowpix is capable of natively processing benchmarks and targeting FPGA implementation of these benchmarks using overlay configurations. This thesis discusses how an application can be expressed using the Flowpix language, and how the compiler processes the application to generate control words that enables the configuration of the hardware architecture. We also compare our results with the existing state-of-the-art frameworks: Polymage, Halide, Xilinx, Darkroom, and Rigel.

Abstract

Biometric authentication plays an increasingly prominent role in today’s products and services for verifying an individual’s identity. It is not only efficient but also practical, as it establishes a unique link to an individual through their physical and behavioral characteristics [43]. Unlike conventional authentication mechanisms like passwords or documents, biometric traits are inherent to each individual, eliminating the need to memorize additional information [64]. However, the security and privacy of biometric templates used in authentication remain primary concerns, as biometric data is strongly and irrevocably tied to an individual, as emphasized in the article [42]. In the context of remote authentication, Secure Multiparty Computation (SMC) offers a powerful solution. SMC enables two parties to interactively compute a function using their private inputs without disclosing any information except for the output itself [19]. This approach ensures that biometric template comparison is carried out in a privacy-preserving manner, enhancing both security and privacy in authentication services. In this thesis, we introduce a unique approach to iris, fingerprint, and face verification by incorporating ”noise” into the authentication process. In our work,“noise” refers to signals obtained from non-discriminatory or unreliable regions of biometric characteristics. Our extensive empirical evaluation reveals a correlation among noise features, and we leverage this correlation in a novel Secure Two-Party Computation (STPC) design. This STPC design operates on quantified uncertainty between noise features, providing informationtheoretic security. Our approach has low accuracy degradations, practical computational complexity, wide applicability making it suitable for practical real-time applications.

Abstract

Location-Based Services (LBSs) have become increasingly prevalent in today’s mobile technology sector, delivering tailored information relevant to the users’ precise locations. These services grant users access to location-centric information like the proximity of hospitals, restaurants, or other points of interest, thereby facilitating routine tasks. However, such LBSs can pose significant concerns about user privacy. Consider a user querying, “What are the directions to the best cancer hospital from the current location?”. Such queries expose the user’s current location information to the LBS provider and other intermediate nodes (intruders) in the mobile network. Query location information can reveal sensitive information about the user, such as relationships, health, religion, and nightlife habits. In this thesis, we propose two improved approaches to preserve the privacy of users’ query location in the mobile environment. As the first approach, we propose an improved dummy generation approach for better privacy. In a dummy generation approach, the user sends additional dummy locations along with the user’s actual location in its query, thereby confusing the LBS provider and the other nodes. The existing approaches have the issue of generating dummies in regions with more infeasible regions (inaccessible regions). Moreover, the existing approaches do not consider the presence of time-dependent infeasible regions. For example, consider a supermarket with opening and closing times as 9am and 9pm, respectively. From 9am to 9pm, this supermarket can be considered a feasible region; otherwise, this area can be regarded as an infeasible region. Furthermore, if the intruder estimated the centre of cloaking region (CR) using the dummy locations, it would become more accessible for the intruder to know a given user’s actual location. To improve the performance, we propose an Annulus-based Gaussian Dummy Generation (AGDG) approach. AGDG introduces the concept of a virtual cloaking region to generate cloaking regions. In AGDG, unlike traditional methods, the user’s location is not fixed at a fixed distance from the centre of the cloaking region. Additionally, AGDG considers the infeasible regions and query probability in the surrounding environment when generating dummy locations. The approach also incorporates the concept of time-dependent infeasible regions and ensures that the generated dummy locations abide by these time-dependent constraints. As the second approach, we propose a cloaking-based approach to improve the privacy of spatial range queries. In distributed spatial cloaking-based approaches, the user’s query location information is cloaked using the distributed mobile network around the user (e.g., the p2p network). Existing approaches do not preserve the user’s intent privacy. For example, suppose a user queries all the cancer hospitals near her. In that case, her location and health information (searching for intent, which is about cancer hospitals) must be preserved from both LBS providers and peers in the surrounding. Moreover, the existing approaches require a large number of peers to be employed to cloak the user query location. Maintaining such structures in a highly dynamic mobile network is challenging. We propose the notion of ijkCloak framework to improve existing distributed spatial cloaking-based approaches. The ijkCloak framework introduces the notion of ijk-anonymity to protect both the user’s query location and intent information. This method divides the user’s query location information into multiple fragmented locations. This process helps keep the user’s query location private from their peers and the LBS provider. Additionally, dummy intents are sent to the LBS provider along with the user’s actual query to protect the user’s query intent. The proposed approach ijkCloak, adopts ijk-anonymity in a mobile network environment. Because of the efficiency of ijk-anonymity, this proposed method requires fewer peers to maintain user privacy, making it more practical in a highly dynamic mobile network environment. For each approach, the theoretical analyses and comprehensive experimental study exhibits its potential to preserve location privacy in different scenarios. We hope this research encourages further research and leads to the development of improved privacy preserving approaches in mobile networks.

Abstract

Player detection is a fundamental building block for numerous applications in sports analytics, encompassing player recognition, player tracking, and activity detection. However, the majority of existing research in this domain relies on fixed-camera top-view videos of the field, which inherently simplifies the player detection task. Regrettably, such videos are not readily accessible to the general public, rendering them an unreliable data source for comprehensive player analysis. In contrast, broadcast videos of matches offer a readily available resource. Performing player detection on these videos proves considerably more challenging due to the presence of diverse sources of noise. This study investigates player detection in the context of continuous long-shot broadcast videos, acknowledging the complexities associated with this particular setting. In the initial phase of our research, we thoroughly examine the distinctions between player detection and person detection while also investigating the multitude of challenges inherent to player detection. We begin by formulating player detection as a domain adaptation problem and analysing the various challenges associated with this approach. Our analysis encompasses an in-depth examination of the overarching challenges encountered in player detection, along with a comprehensive exploration of the unique obstacles posed specifically by broadcast videos and domain adaptation settings. During the subsequent phase of our research, we worked on the development of an extensively annotated player detection dataset, curated from soccer broadcast videos of the FIFA 2018 World Cup matches. This dataset serves as a robust foundation for evaluating the efficacy of player detection algorithms within the context of broadcast videos. We devise a comprehensive pipeline for generating automatic labels for our dataset, which are then corrected further down the pipeline to facilitate the annotation process. The resultant dataset comprises over 200,000 high-resolution frame images, encompassing more than 2,000,000 annotated bounding boxes extracted from three distinct FIFA 2018 World Cup matches. Notably, our dataset encompasses a diverse set of player positions, orientations, and bounding box sizes, effectively capturing the inherent variability encountered in soccer broadcasts. Additionally, the dataset incorporates numerous instances of challenging noisy data points, elevating its complexity beyond previous datasets in the field. In the third phase of our research, we present a novel transductive approach to address the player detection challenge, treating it as a domain adaptation problem. We demonstrate the significance of instance-level domain labels in achieving effective adaptation, specifically for soccer broadcast videos. To efficiently annotate these domain labels on the bounding box predictions generated by our inductive model, we propose a sophisticated multi-model greedy labelling scheme that leverages visual features. The annotated domain labels are then utilized to train a transductive counterpart of the model, utilizing reliable instances derived from the inductive model inferences. This approach proves to be highly advantageous, enabling remarkable performance enhancements for a given match with a minimal number of labelled samples. Our experimental results highlight an average increase of 16 points in mean Average Precision (mAP) for soccer broadcast videos, accomplished by annotating domain labels for approximately 100 samples per video. In the culminating phase of our research, we demonstrate the practical utilization of robust player detection algorithms in constructing analytical systems to enhance game analysis. Specifically, we develop field heat maps that effectively depict the spatial distribution of players on the field over time. Leveraging bounding box detections derived from our proposed approach, we employ homographic projections to achieve accurate top-view registration of the detected bounding boxes in each frame. These generated heat maps serve as a valuable resource for deriving insightful inferences that directly correlate with significant events transpiring during the match. Furthermore, we present additional potential applications for leveraging reliable detection systems, while also outlining avenues for future enhancements and refinements to our system.

Abstract

In the present era of Big Data, the demand for storing vast amounts of data is rapidly increasing among companies such as Facebook, Microsoft, Google, Intel, IBM, and others, for various applications. To address this need, Distributed Storage Systems (DSSs) have been established, offering improved capabilities in terms of flexibility, scalability, speed, and cost. In DSS, data is distributed and stored on different nodes and are connected through the network. However, data loss is inevitable due to physical limitations such as hardware failures and power shutdowns. Maximum Distance Separable (MDS) codes are very efficient in terms of storage overhead. For practicality, Locally Recoverable codes (LRCs) are discovered to facilitate the low reconstruction cost for single and multiple failures (Independent and correlated), with a slight increase in storage overhead. Maximally recoverable codes are a class of codes that recover from all potentially recoverable erasure patterns given the locality constraints of the code. Our main objectives are to provide MRCs for independent failures, correlated failures with low computational complexity, and encoding complexity. In earlier works, codes have been studied in the context of codes with the locality to handle independent failures. The notion of locality has been extended to the hierarchical locality, which allows for the locality to gradually increase in levels with the increase in the number of erasures. In one direction, we consider MRC for the case of codes with 2-level hierarchical locality for the specific topology (locality constraints) called Hierarchical Local MRC (HLMRC). We derive a field size lower bound on HL-MRC. We also give constructions of HLMRC for some parameters whose field size is smaller than that of earlier known constructions. We investigate Locally Recoverable Codes (LRCs) with availability, which refers to the ability to have multiple repair sets. The presence of multiple repair sets in LRCs is beneficial as it facilitates the distribution of the repair load among various nodes. This distribution helps prevent excessive strain on specific nodes and promotes a more balanced workload within the system. In our research, we expand on the concept of availability in Locally Recoverable Codes (LRCs) and apply it to codes with hierarchical locality. The minimum distance plays a vital role in determining the codes’ capability to handle erasures effectively. Our study focuses on investigating the upper bound of the minimum distance for the specific case of LRCs with hierarchical locality. To reduce the encoding complexity, Halbawi et al. introduced sparse and balanced generator matrices for MDS (Reed - Solomon) code and LRCs (Tamo-Barg code) for single erasure. Building upon this work, we contribute by presenting sparse generator matrices for MRC with locality for single erasure. Furthermore, we also provide sparse and balanced generator matrices for MRC with locality, specifically for single erasures where the locality value is set to 2. In order to deal with correlated failures, Gopalan et al. initiated the study of MRCs gridlike and product topologies. In another research direction, we focus on MR codes for product topology Tm,n(a, b). Product codes are a class of codes with generator matrices as the tensor product of the generator matrices of component codes. The codeword can be represented as an m × n array, where the component codes are referred to as the row and column codes. We derive a few properties of maximally recoverable product codes. We give a sufficient condition to characterize a certain subclass of erasure patterns as correctable and another necessary condition to characterize another subclass of erasure patterns as not correctable. We construct a certain bipartite graph based on the erasure pattern satisfying the regularity condition (a necessary condition for recoverability) for product topology and show that there exists a complete matching in this graph. We used this condition to identify a subset of recoverable erasure patterns for a = 2. In earlier work, higher-order MDS codes denoted by MDS(l) have been defined in terms of generic matrices, and these codes have been shown to be constituent row codes for maximally recoverable product codes for the case of a = 1. We derive a certain inclusion-exclusion type principle for characterizing the dimension of intersection spaces of generic matrices. Applying this, we formally derive a relation between MDS(3) codes and points/lines of the associated projective space