ISI Digital Commons (Indian Statistical Institute )
Not a member yet
7571 research outputs found
Sort by
Essays in Industrial Organization
No full textThis thesis contains three chapters on industrial organization, covering two distinct topics. Chapters 1 and 2 examine the impact of firm owner\u27s corruption on the organization, while Chapter 3 analyzes the market dynamics of a two-sided newspaper market.\medskip The first chapter investigates the agency conflict that arises when the principal (the owner) of a firm is involved in corruption. A corruptible principal has an incentive to conceal his or her illegal activities, while the agent (CEO or manager), due to their information advantage, is in a position to monitor this corruption, thus creating an agency conflict. We show that such corruption leads to increased bureaucracy within the firm as the principal reduces information flow, provides lower incentive wages, and limits delegation to the manager. Furthermore, we analyze additional inefficiencies caused by such corruption, including the principal\u27s incentive to distort talent by hiring a corruptible manager and to expropriate from minority shareholders.\medskip The second chapter investigates a screening mechanism through which a corruptible principal screens a manager when the manager\u27s type (honest or corruptible) is private information at the time of hiring, and the corruptible manager can misappropriate funds from the firm. Our results show that if the potential for misappropriation by the manager is within an intermediate range, the principal can design a wage contract that ensures only a corrupt manager joins. This range increases if the reservation wage rises. We also demonstrate that the principal can completely offset the cost of the manager\u27s misappropriation through a suitable wage contract, if these costs are not critically high.\medskip The third chapter extends the vertical differential framework by incorporating the advertisement side to analyze the two-sided newspaper market. Newspaper markets are highly concentrated, with most being monopolies or duopolies within a service area. Existing literature attributes this market concentration to the network effect, which arises because newspaper readers derive positive utility from advertisements, especially classifieds. We demonstrate that newspaper markets can also be concentrated due to endogenous investment in quality, particularly when quality improvements involve fixed costs like newsroom size. This reason more closely aligns with empirical evidence, which shows that market concentration persists even when classified ad revenues declined significantly due to online platforms like Craigslist. We also show that several different types of market and product configurations emerge depending on advertisement levels
Unitary Connections and Q-Systems
No full textIn the classification programme of subfactors, the standard invariant plays a quintessential role. In this thesis, we furnish a 2-categorical perspective of the standard invariant of finite-index subfactors and further study algebraic structures (Q-systems) associated with them
A classical and simulated approach of cost optimisation in reverse logistics process for e-commerce business
No full textThe purpose of this paper is to establish the proper utilisation and allocation of resources in an e-commerce reverse logistics network to minimise the overall cost. This study formulates the reverse logistics network of e-commerce business into a binary integer programming model. It solves a numerical problem by this classical approach to find out the optimal cost through optimal assignment of resources and path allocation for transporting the products. Simulation is performed next to replicate the real-life reverse logistics scenario of e-commerce business. Finally, a sensitivity analysis is carried out to observe the impact of proper utilisation of the resources on reducing the total operation cost significantly. It encourages e-commerce business for optimal allocation of logistics agents and vehicles, and utilisation of the regional hubs and sorting facilities for transporting the returned products. The study also highlights the scope of future research both from theoretical and managerial perspectives
A new deep CNN for 3D text localization in the wild through shadow removal
No full textText localization in the wild is challenging due to the presence of 2D and 3D texts, the presence of shadows, arbitrary orientated text with non-linear arrangements, varying lighting conditions as well as complex background. This paper proposes the first approach for 3D text localization in natural scene images through shadow removal and a new deep CNN model. In a first step, exploiting the observation that 3D text generates shadow information in natural scenes, the proposed model detects and removes the shadow pixels of 3D text based on the Generalized Gradient Vector Flow concept and a new clustering approach. The performance of the classification of 2D and 3D texts in the scene images is strengthened by using key features, including pixel strength, sharpness and edge potential, which are extracted to eliminate false text and shadow pixels. For text localization after removing shadow information, EfficientNet is used as an encoder (backbone) and UNet as a decoder in a novel way employing differential binarization. Experimental validation and comparative analysis with state-of-the-art approaches on both a new purpose-built dataset as well as on the benchmark datasets of ICDAR MLT 2019, ICDAR ArT 2019, CTW1500, DAST1500, Total-Text, and MSRATD500 for each of the different steps of the method, show that the proposed approach outperforms the existing methods
Sample and Query Complexities of Some Estimation Problems
No full textGiven data from some experiment, inferring information from the underlying distribution is of prime importance, and has been extensively studied. However, due to the huge size of the data, traditional methods are often no longer applicable. Thus new tools and techniques are being developed for inferring useful information from large amounts of data. This thesis makes progress in this direction. The primary goal is to design efficient randomized algorithms aka. testers that can distinguish whether a given unknown object is close\u27\u27 or far\u27\u27 from a property of interest with as few accesses as possible. This is referred to as distribution testing when the unknown object is a probability distribution, and graph property testing when it is a graph. The minimum number of samples required to decide a property in distribution testing is referred to as sample complexity, while in graph property testing, it is referred to as query complexity. In this thesis, we study several fundamental problems in distribution and graph property testing such as (i) Can one design a tolerant tester for any distribution property with only black-box access to a non-tolerant tester? (ii) Does there exist distribution properties with global structure that can be learnt efficiently? (iii) the role of adaptivity in distribution testing, and tolerant testing for (iv) graph isomorphism and (v) bipartiteness. The results of the thesis are divided into three parts. In the first part, we study the connection between the sample complexities of non-tolerant and tolerant testing of distributions and prove a tight quadratic gap for label-invariant (symmetric) properties, while providing lower bounds for non-concentrated properties. We also present an algorithm that can learn a concentrated distribution even when its support set is unknown apriori. In the second part, we investigate problems (ii) and (iii) in huge object model, where distributions are defined over n-dimensional Hamming cube and the tester obtains n-bit strings as samples. Since reading the string in its entirety may not be feasible for large n, the tester has query access to the sampled strings. We define the notion of index-invariant properties, properties that are invariant under the permutations of the indices {1,......,n} and prove that any index-invariant property whose VC-dimension is bounded has a tester whose query complexity is independent of n and depends only on VC-dimension. Moreover, the dependencies of sample and query complexities of our tester on the VC-dimension are tight. We also study the power of adaptiveness in this model and prove a tight quadratic separation between query complexities of adaptive and non-adaptive testers for index-invariant properties, compared to tight exponential separation for its non-index-invariant counterpart. In the third part, we study property testing of dense graphs and give positive answers to problems (iv) and (v). We prove that tolerant graph isomorphism testing is equivalent to the problem of estimating the Earth Mover Distance of two distributions, constructed from the graphs. Moreover, our equivalence proof is model independent. Finally we design a tester for tolerant bipartiteness testing whose query and time complexities are significantly better compared to previous works
A flexible Bayesian variable selection approach for modeling interval data
No full textInterval datasets are not uncommon in many disciplines including medical experiments, econometric studies, environmental studies etc. For modeling interval data traditionally separate models are used for modeling the center and the radius of the response variable. In this article, we consider a Bayesian regression framework for jointly modeling the center and the radius of the intervals corresponding to the response, and then use appropriate priors for variable selection. Unlike the traditional setting, both the centres and the radii of all the predictors are used for modeling the center and the radius of response. We consider spike and slab priors for the regression coefficients corresponding to the centers (radii) of the predictors while modeling the center (radius) of the response, and global–local shrinkage prior for the coefficients corresponding to the radii (centers) of the predictors. Through extensive simulation studies, we illustrate the effectiveness of our proposed variable selection approach for the analysis and prediction of interval datasets. Finally, we analyze a real dataset from a clinical trial related to the Acute Lymphocytic Leukemia (ALL), and then select the important set of predictors for modeling the lymphocyte count which is an important biomarker for ALL. Our numerical studies show that the proposed approach is efficient, and it provides a powerful statistical inference for handling interval datasets
A New Contrastive Learning-Based Vision Transformer for Sentiment Analysis Using Scene Text Images
No full textSentiment analysis using scene text images is complex and challenging because it has an arbitrary background, and the method should rely on only visual features. Unlike most existing methods that use either text or images or both, this study uses only scene text images for sentiment analysis. The intuition to use only scene text images is that sometimes users express their feelings and emotions or convey their messages by writing text in different shapes with diverse background designs. It is noted that the existing methods ignore such vital cues for sentiment analysis. This work explores a vision transformer to extract visual features that represent contextual information about the appearance of the text image. Further, to strengthen the visual features, the proposed work introduces contrastive learning which maximizes the gap between inter-classes and minimizes the gap between intra-classes of positive, negative, and neutral. To demonstrate the effectiveness of the proposed method, it is tested on our own constructed dataset and benchmark dataset. A comparative study of our method with the existing method shows the proposed method is superior in the classification of positive, negative, and neutral scene text images
A new U-Net based system for multi-cultural wedding image classification
No full textUse of social media for communication, sharing or expressing views, broadcasting news, threatening and blackmailing has become an integral part of society. One such activity is understanding multi-cultural wedding images uploaded on social media. This paper presents a novel method based on the combination of U-Net, Convolutional Neural Network and Random Forest for classification of multicultural wedding images. In the case of wedding images, bride and bridegroom draw the attention of the viewers. This observation led to propose a U-Net for segmenting the region of bride and bridegroom in a novel way. Similarly, it is noted that the costumes of bride and bridegroom are vital information for differentiating different cultures. This cue motivated us to extract features using CNN for classification. Since the extracted features using CNN are capable of discriminating images of different classes, we propose a simple and effective Random-Forest for Multicultural Wedding Image Classification. The efficiency of the proposed model is demonstrated by testing it on our own dataset of six multi-cultural wedding classes and standard dataset of wedding and non-wedding images classes. Experimental results on both the datasets show that the proposed model outperforms the state-of-the-art models in terms of average classification rate
A note on an invariant distance of the bidisk
No full textIn this short paper, we discuss relation between an invariant distance of the bidisk and Kreĭn space geometry. In particular, an interpolation theorem for rational maps with respect to our invariant distance is proven
A review on association between menopausal symptoms and cardiovascular risk factors
No full textMenopausal transition and post-menopausal periods can have short-term and long- term effects on mid-life health of women. The short-term effects include the possibility of experiencing of menopausal symptoms, while the long-term effects include cardiovascular diseases (CVD) risk. The occurrence of menopausal symptoms varies widely within and between populations. Studies indicate that the frequency and severity of menopausal symptoms are linked to CVD risk factors, but the existing literature is divergent and somewhat limited. Thus, women belonging to different populations are likely to be at a different risk of CVD, but the exact physiological mechanism behind this relationship remains unclear. The present narrative review aimed to synthesize the available evidence of menopausal symptoms in association with various conventional CVD risk factors such as blood pressure, total cholesterol and blood glucose levels and obesity, as well as to determine the potential link between these two processes. We undertook a rigorous data base search to identify, examine, and critically assess the existing literature on the associations between menopausal symptoms and CVD risk factors. We applied inclusion and exclusion criteria to filter the retrieved articles and classified the literature into eight major categories. The risk of CVD is higher among women who experience vasomotor, psychological, and urogenital symptoms compared to those who do not experience these symptoms. Our review indicates that menopausal symptoms can be used as markers in assessing CVD risk factors during midlife. Thus there is a need for larger-scale research to support these findings and identify the potential mediators that are controlling this association