ISI Digital Commons (Indian Statistical Institute )
Not a member yet
7571 research outputs found
Sort by
Machine Learning based Decoding of Heavy Hexagonal QECC for Asymmetric Quantum Noise
No full textDecoding error syndromes for topological quantum error correcting codes, such as surface and heavy hexagonal codes, is computationally expensive. While minimum weight perfect matching (MWPM) algorithms have been commonly used for decoding, recent works have demonstrated the efficacy of machine learning (ML), particularly neural networks, in decoding syndromes for these codes. In this study, we introduce a ML-based decoder tailored to heavy hexagonal code to address asymmetric noise channels which reflect real-world scenario better than the depolarization model considered in previous works. Our proposed decoder shows \sim 5× and Isim 22× improvements in the threshold values for amplitude and amplitude-phase damping noise models respectively over MWPM methods. Our decoder is also robust to changes in asymmetry, with the threshold reducing by only 3.6% for a 10× change in asymmetry
Maximal Independent Set via Mobile Agents
No full textWe consider the problem of finding a maximal independent set (MIS) in an unknown graph by mobile agents or mobile robots. Suppose n mobile agents are initially located at arbitrary nodes of an n-node anonymous graph G = (V, E) and the goal is that the agents autonomously relocate themselves to find a subset S gV of nodes such that S forms an MIS. The objective is to minimize both (i) the time required to find an MIS and (ii) the memory required at each agent. We consider the mobile agents with communicate-compute-move model in the synchronous setting, in which agents can communicate with other agents if they are located at the same node (called local communication). We present deterministic algorithms for finding MIS on various graph classes (e.g., general graphs, trees, grids) in our robot model. Additionally, we consider different initial configurations of the agents over the nodes. Specifically, we present algorithms for three different initial configurations: rooted (all agents are positioned at the same node), dispersed (one agent at each node), and arbitrary (agents positioned at multiple nodes, except in the dispersed configuration). For general graphs, our algorithm finds MIS in O(nI) time and uses O(log n) bits of memory per agent in the rooted initial configuration, where Δrepresents the maximum degree of the graph. The algorithms for the dispersed and arbitrary initial configurations require O(nIlog n) time and O(log n) bits of memory, but necessitate prior knowledge of n and I. For trees, our algorithms find MIS in: (i) O(D2) time and O(log n) bits of memory per agent in the rooted configuration, where D is the diameter of the tree, (ii) O(D) time and O(Δ+ log n) bits of memory per agent in the dispersed configuration, requiring the knowledge of I, and (iii) O(n) time and O(Δ+ log n) bits of memory per agent in the arbitrary configuration, where knowledge of both n and Δis required. Lastly, for rectangular grids with n = x × y nodes, our algorithm finds MIS in O(max (x, y)) time and O(log n) bits memory per agent, requiring the knowledge of x, y. The algorithm is simultaneously time and memory optimal. To the best of our knowledge, this is the first time where the maximal independent set problem is explored through the lens of mobile agents in the local communication model
Pathologist-Like Explanations Unveiled: An Explainable Deep Learning System for White Blood Cell Classification
No full textDespite of achieving remarkable accuracy, the capability of deep learning models for robust prediction of explanations remains largely unexplored in white blood cells (WBCs) classification. In this study, we introduce HemaX, an explainable deep neural network-based model that produces pathologist-like explanations using five attributes: granularity, cytoplasm color, nucleus shape, size relative to red blood cells, and nucleus to cytoplasm ratio (N:C), along with cell classification, localization, and segmentation. HemaX is trained and evaluated on a novel dataset, LeukoX, comprising 467 blood smear images encompassing ten (10) WBC types. The proposed model achieves impressive results, with an average classification accuracy of 81.08% and a Jaccard index of 89.16% for cell localization. HemaX successfully predicts the five explanations with a normalized mean square error of 0.0317 for N:C ratio and over 80% accuracy for the other four attributes. Through expert validations and multiple empirical analyses, we illustrate the robustness of HemaX towards both cell classification and explanation prediction
Poincaré and Picard bundles on the moduli spaces of bundles on curves
No full textPoincare bundles and Picard bundles on the Jacobian of a smooth curve have been studied extensively classically. Peter Newstead and his collaborators have made major contributions to the theory of Poincar´e and Picard bundles over the moduli spaces of vector bundles of higher ranks. We present old and new results on the Poincare bundles and Picard bundles on the moduli spaces of vector bundles on smooth curves and their generalisations to nodal curves. Finally we consider parabolic vector bundles. We prove the parabolic stability of Poincar´e bundles over the moduli spaces of parabolic vector bundles, with parabolic structure of any type, on a smooth curve
PReLim: A Modeling Paradigm for Remote Sensing Image Scene Classification Under Limited Labeled Samples
No full textWith the ongoing development of deep learning techniques in recent years, the convolutional neural networks (CNNs) have shown remarkable performance breakthrough in remote sensing image scene classification. However, the performance of these deep models largely depends on the number of available training samples or labeled images. Although the knowledge transferring and pre-training techniques can handle such situation, these may become ineffective due to domain difference. On the other side, the existing data augmentation approaches often produce training samples with too low diversity to help in performance improvement. In order to address these issues, in this work, we propose PReLim as a novel modeling paradigm for remote sensing scene classification under limited labeled samples scenario. PReLim is based on the notion of local and global filtering of scene fragment mixture, which overcomes both the sample diversity and the domain difference issue. Experimental analyses with the benchmark UCMerced and SIRI-WHU datasets demonstrate the effectiveness of PReLim in achieving the state-of-the-art accuracy using limited number of training samples
Rough Algebraic Semantics of Concepts in a Distributed Cognition Perspective
No full textUp-directed rough sets are introduced and studied by the present author in earlier papers. This is extended by her in two different granular directions in this research, with a surprising algebraic semantics. The granules are based on ideas of generalized closure under up-directedness that may be read as a form of weak consequence. This yields approximation operators that satisfy cautious monotony, while pi-groupoidal approximations (that additionally involve strategic choice and algebraic operators) have nicer properties. The study is primarily motivated by possible structure of concepts in distributed cognition perspectives, real or virtual classroom learning contexts, and student-centric teaching. This study thus provides directions for building AI models of distributed cognition, and related decision-making in general
Evolving Nature of Women Empowerment in India
No full textThis chapter takes a fresh look at the issue of women\u27s empowerment and its relationship with development. The authors address the slower pace of development in women empowerment in India by looking at two aspects: empowerment and determinants. In this regard, they construct a measure of women empowerment (WEMP) onboarding its multiple dimensions such as education, health, domestic autonomy, and access to free mobility-related indicators. Second, they analyze the association of different socioeconomic, demographic, and household-level characteristics with the women empowerment index measure. In this regard, they exploit multiple rounds of National Family Health Survey (NFHS) data – the NFHS-IV(2015–16), and the NFHSV (2019–20) at the district level. Our results show that the districts in the southern region not only perform better in terms of empowering women but also remain in the top-quintile of the distribution of the WEMP index. By comparison, poor-performing districts in the central-eastern region have made certain improvements during this period, although they remain in the lower quintiles. Our multivariate regressions for two rounds separately indicate the dominance of family-level demographics and educational and family structure variables in explaining WEMP as compared to socioeconomic and sociodemographic variables
Maternal and Child Nutritional Status in South and Southeast Asia: A Comparative Analysis Across Bangladesh, Cambodia, India, Maldives, Nepal, Pakistan, and Timor-Leste
No full textObjectives: This research aims to investigate the impact of maternal characteristics on the nutritional status of under-five children in South and Southeast Asian countries using nationally representative data. Setting: The study was conducted using nationally representative data collected from seven countries, namely, Bangladesh, Cambodia, India, Maldives, Nepal, Pakistan, and Timor-Leste. The data was collected between the years 2016 and 2022. The study included 2, 17, 045 mother–child pairs. Results: The study reveals a significant association between maternal attributes and various markers of child development. It suggests that maternal age, height, weight, and BMI are closely linked to the height, weight, HAZ, WAZ, WHZ, and BAZ of children. Mothers in Maldives have the highest body weight, falling under the overweight category, while mothers in Timor-Leste have the lowest weight and BMI. This diversity in maternal characteristics emphasizes the complexity of the issue and its potential impact on their child\u27s development. It is worth noting that boys and girls in the Maldives and Cambodia have the highest height and weight, while the lowest child weight is found in Timor-Leste, and the lowest height is found in Pakistan.Conclusion: The findings of the study indicate that a child\u27s height and weight can be influenced by their mother\u27s age, height, weight, and BMI. The study also revealed that Timor-Leste is the most vulnerable country in terms of childhood malnutrition in the South and Southeast Asian region
Action Detection System for Dark Videos using Spatio-Temporal Features and Bidirectional Encoder Representations from Transformers
No full textWeighted inequalities for maximal operators and the Hardy space H1 on LCA groups
No full textThe purpose of this thesis is two fold: to study weighted norm inequalities for maximal type operators such as Hardy{Littlewood maximal operator associated with a family of general sets in a topological space and Fourier maximal operator in the context of the ring of integers of a local eld, and to extend the classical theory of Hardy space and related topics such as the space BMO and the John{Nirenberg space in the setting of Locally Compact Abelian (LCA) groups having a covering family. In chapter 2 we study norm inequalities for the maximal operator ME associated with a family E of general sets from various points of view. Our rst main result is the mixed Ap �� A1 weighted estimates for the operator ME. The main ingredient to prove this result is a sharp form of a weak reverse H older inequality for the A1;E weights. As an application of this inequality, we also provide a quantitative version of the open property for Ap;E weights. Our second main result in this setting is the establishment of the endpoint Fe erman{Stein weighted inequalities for the operator ME. Furthermore, vector-valued extensions for maximal inequalities are also obtained in this context. Chapter 3 focuses on the weighted norm inequalities for Fourier series in the context of the ring of integers D of a local eld K and some important applications. We establish weighted estimates for the maximal partial sum operator M of Fourier series on the weighted spaces Lp(D;w), 1 \u3c p \u3c 1, where w is a Muckenhoupt Ap weight. As a consequence of this result, we obtain the uniform boundedness of the Fourier partial sum operators Sn; n 2 N, on Lp(D;w). Both these results include the cases when D is the ring of integers of the p-adic eld Qp and the eld Fq((X)) of formal Laurent series over a nite eld Fq, and in particular, when D is the Walsh{Paley or dyadic group 2!. The aim of this chapter 4 is to extend the classical theory of the Hardy space H1 and its dual space of BMO functions with \bounded mean oscillation to the setting of LCA groups G having covering families. First, we discuss in details the setting of LCA groups where our work is developed. Next, we introduce the notion of atomic Hardy spaces H1;q(G) with atom parameter 1 \u3c q 1 and the notion of the space BMO(G) in this setting. After presenting some basic properties of these spaces, we then establish the main feature for functions in BMO(G), namely the John{Nirenberg inequality. Moreover, we show that the atomic Hardy spaces H1;q(G) are independent of the choice of the parameter q. Finally, we relate H1;q(G) with BMO(G) via duality in this setting. Finally chapter 5 of this thesis explores the theory of John{Nirenberg spaces JNp in the setting of LCA groups having covering families. The main result of this chapter is the John{Nirenberg inequality for functions in JNp spaces which describes, as it happens in Euclidean setting, that JNp can be embedded into weak Lp spaces