ISI Digital Commons (Indian Statistical Institute )
Not a member yet
7571 research outputs found
Sort by
Testing distributional equality for functional random variables
No full textIn this article, we present a nonparametric method for the general two-sample problem involving functional random variables modeled as elements of a separable Hilbert space H. First, we present a general recipe based on linear projections to construct a measure of dissimilarity between two probability distributions on H. In particular, we consider a measure based on the energy statistic and present some of its nice theoretical properties. A plug-in estimator of this measure is used as the test statistic to construct a general two-sample test. Large sample distribution of this statistic is derived both under null and alternative hypotheses. However, since the quantiles of the limiting null distribution are analytically intractable, the test is calibrated using the permutation method. We prove the large sample consistency of the resulting permutation test under fairly general assumptions. We also study the efficiency of the proposed test by establishing a new local asymptotic normality result for functional random variables. Using that result, we derive the asymptotic distribution of the permuted test statistic and the asymptotic power of the permutation test under local contiguous alternatives. This establishes that the permutation test is statistically efficient in the Pitman sense. Extensive simulation studies are carried out and a real data set is analyzed to compare the performance of our proposed test with some state-of-the-art methods
The eco-evolutionary dynamics of strategic species
No full textMuch research has in recent years been devoted to better our understanding of the intricate relationships between ecology and the evolutionary success of species. These explorations have often focused on understanding the complex interplay among ecological factors and evolutionary rhythms of the species in various environments. Central to these studies is the concept of the survival of the fittest, proposed by Charles Darwin, where evolutionary circumstances, often portrayed as social dilemmas, favour the welfare of self-interested over others. To further advance this line of research, we here develop a theoretical framework that features thre
The Extended Bregman Divergence and Parametric Estimation in Continuous Models
No full textUnder standard regularity conditions, the maximum likelihood estimator (MLE) is the most efficient estimator at the model. However, modern practice recognizes that it is rare for the hypothesized model to hold exactly, and small departures from it are never entirely unexpected. But classical estimators like the MLE are extremely sensitive to the presence of noise in the data. Within the class of robust estimators, which constitutes parametric inference techniques designed to overcome the problems due to model misspecification and noise, minimum distance estimators have become quite popular in recent times. In particular, density-based distances under the umbrella of the Bregman divergence have been demonstrated to have several advantages. Here we will consider an extension of the ordinary Bregman divergence, and investigate the scope of parametric estimation under continuous models using this extended divergence proposal. Many of our illustrations will be based on the GSB divergence, a particular member of the extended Bregman divergence, which appears to hold high promise within the robustness area. To establish the usefulness of the proposed minimum distance estimation procedure, we will provide detailed theoretical investigations followed by substantial numerical verifications
A Contrastive Explanation Tool for Plans in Hybrid Domains
No full textThis paper presents a tool for generating contrastive explanations of plans in hybrid domains. The tool offers a collection of contrastive questions over a plan, such as “Why use action A?”, for users to select. An explanation of the user question is produced by contrasting the original plan against an alternative that meets the user’s expectation implicit from the question. The tool has the provision to contrast with the best alternative that the underlying planner can generate in terms of plan makespan (duration of a plan) and length. The current version supports two planners in the hybrid domain, namely SMTPlan+, and ENHSP, which a tool user can select. The tool consists of (1) A web-based interactive GUI for selecting questions, viewing contrastive plans and the generated explanations, and (2) A back-end implementing an iterative re-modeling and re-planning algorithm. We demonstrate the working of the tool over two case studies
Adverse Drug Event Prediction with a Multi-Layer Heterogeneous Graph Neural Network Architecture
No full textDrug-drug interactions (DDIs) and their associated adverse drug effects (ADEs) pose significant challenges in polypharmacy, particularly for patients with complex or co-occurring conditions. Existing methods often struggle to capture the intricate relationships between drugs, proteins, and the problem of multiple ADEs between two drug interactions. We introduce MI-GNN, a novel multi-layered graph neural network framework designed to address the multiple ADE problem between two drugs while capturing the dependencies between drug and target proteins for predicting DDIs. Our approach models the adverse events as multiple layers, allowing information flow between layers to capture inter-dependencies among ADEs. We validated MI-GNN\u27s efficacy by performing ADE prediction tasks on the benchmark dataset while comparing it with state-of-the-art prediction algorithms. Experimental results demonstrate that MI-GNN effectively predicts ADEs while incorporating knowledge gained from drug and protein embeddings. Our proposed model outperforms the baseline established in previous work by more than 2%. MI-GNN offers a promising approach for modeling complex polypharmacy side effects by leveraging multi-modal graph structures and capturing interdependencies among drugs, proteins, and ADEs
Approximate Degree Composition for Recursive Functions
No full textDetermining the approximate degree composition for Boolean functions remains a significant unsolved problem in Boolean function complexity. In recent decades, researchers have concentrated on proving that approximate degree composes for special types of inner and outer functions. An important and extensively studied class of functions are the recursive functions, i.e. functions obtained by composing a base function with itself a number of times. Let hd denote the standard d-fold composition of the base function h. The main result of this work is to show that the approximate degree composes if either of the following conditions holds: The outer function f : {0, 1}n → {0, 1} is a recursive function of the form hd, with h being any base function and d = Ω(log log n). The inner function is a recursive function of the form hd, with h being any constant arity base function (other than AND and OR) and d = Ω(log log n), where n is the arity of the outer function. In terms of proof techniques, we first observe that the lower bound for composition can be obtained by introducing majority in between the inner and the outer functions. We then show that majority can be efficiently eliminated if the inner or outer function is a recursive function
Free Lunch: Frame-level Contrastive Learning with Text Perceiver for Robust Scene Text Recognition in Lightweight Models
No full textLightweight models play an important role in real-life applications, especially in the recent mobile device era. However, due to limited network scale and low-quality images, the performance of lightweight models on Scene Text Recognition (STR) tasks is still much to be improved. Recently, contrastive learning has shown its power in many areas, with promising performances without additional computational cost. Based on these observations, we propose a new efficient and effective frame-level contrastive learning (FLCL) framework for lightweight STR models. The FLCL framework consists of a backbone to extract basic features, a Text Perceiver Module (TPM) to focus on text-relevant representations, and a FLCL loss to update the network. The backbone can be any feature extraction architecture. The TPM is an innovative Mamba-based structure that is designed to suppress features irrelevant to the text content from the backbone. Unlike existing word-level contrastive learning, we look into the nature of the STR task and propose the frame-level contrastive learning loss, which can work well with the famous Connectionist Temporal Classification loss. We conduct experiments on six well-known STR benchmarks as well as a new low-quality dataset. Compared to vanilla contrastive learning and other non-parameter methods, the FLCL framework significantly outperforms others on all datasets, especially the low-quality dataset. In addition, character feature visualization demonstrates that the proposed method can yield more discriminative character features for visually similar characters, which also substantiates the efficacy of the proposed methods. Codes and the low-quality dataset will be available soon
On the Asymmetry of Stuck-at Fault Sensitivity in Memristive Neural Architectures
No full textThe use of memristive crossbar-based architectures has gained traction as a potential solution for performing computationally expensive tasks such as matrix-vector multiplication and vector outer product, which require significant amount of space, time, and energy. Despite being deemed inherently fault-tolerant, memristive crossbar-based neural architecture (MCNA) may often experience accuracy degradation due to hardware faults, resulting in significant variations. This study aims to comprehensively analyze the impact of stuck-at faults (SAFs) on the accuracy of a neural network during classification or regression. Contrary to the popular belief, it is observed that the impact of stuck-at-0 (SAO) and stuck-at-l (SAl) faults are highly asymmetric with respect to the loss of accuracy. Thus this study might help in planning test strategies for the enhancement of fault immunity in memristive neural architectures
Tight Security of TNT and Beyond
No full textLiskov, Rivest and Wagner laid the theoretical foundations for tweakable block ciphers (TBC). In a seminal paper, they proposed two (up to) birthday-bound secure design strategies — LRW1 and LRW2 — to convert any block cipher into a TBC. Several of the follow-up works consider cascading of LRW-type TBCs to construct beyond-the-birthday bound (BBB) secure TBCs. Landecker et al. demonstrated that just two-round cascading of LRW2 can already give a BBB security. Bao et al. undertook a similar exercise in context of LRW1 with TNT — a three-round cascading of LRW1 — that has been shown to achieve BBB security as well. In this paper, we present a CCA distinguisher on TNT that achieves a non-negligible advantage with O(2n/2) queries, directly contradicting the security claims made by the designers. We provide a rigorous and complete advantage calculation coupled with experimental verification that further support our claim. Next, we provide new and simple proofs of birthday-bound CCA security for both TNT and its single-key variant, which confirm the tightness of our attack. Furthering on to a more positive note, we show that adding just one more block cipher call, referred as 4-LRW1, does not just re-establish the BBB security, but also amplifies it up to 23n/4 queries. As a side-effect of this endeavour, we propose a new abstraction of the cascaded LRW-design philosophy, referred to as the LRW+ paradigm, comprising two block cipher calls sandwiched between a pair of tweakable universal hashes. This helps us to provide a modular proof covering all cascaded LRW constructions with at least 2 rounds, including 4-LRW1, and its more established relative, the well-known CLRW2, or more aptly, 2-LRW2
YOLO Assisted A* Algorithm for Robust Line Segmentation of Degraded Document Images
No full textAlthough OCR from images of good quality documents can be considered as a solved problem, the same is not true whenever its quality gets degraded due to certain reasons such as its very old age. On the other hand, OCR of old documents has significant importance towards preservation of cultural heritage, indexing, retrieval etc. The task of degraded document OCR is often critical due to a number of reasons, including the high resemblance between noisy background and faded foreground pixels, asymmetric skews of different lines etc. The study presented in this article has been conducted on a dataset of recently collected sample images of old severely degraded document pages in addition to a few others and the task is very difficult due to the high degradation level of the samples and lack of training ground truths. Here, we propose a hybrid approach combining both of a learning-based and another rule-based methods for line segmentation of similar degraded documents. The proposed method utilizes well-known object detection system YOLO, trained on a publicly available dataset of handwritten samples, to predict starting point (left extreme point) of each line divider, the remaining part of the segmenting line has been obtained using a modified version of graph traversing approach ‘A* path finding’. Thus, the path of the segmenting line suitably dividing two consecutive text lines starting from the predicted left end point and terminating at the right end point could be obtained. The proposed approach has overcome various existing challenges of line segmentation of old degraded quality documents and improved results on several publicly available datasets. Performance comparisons of three existing strategies on five datasets of different languages and varying degradation levels, both of printed and handwritten texts have been presented in this article