Abstract

As we all know, training a highly effective online learning model for complex tasks often hinges on the abundance of high-quality noise-free labeled data. However, acquiring this high-quality, noise-free labeled data is becoming a bottleneck in cost, time, and computational resources. In this thesis, we make a sincere effort to address two significant issues that the current State-Of-TheArt bandit feedback-based online learning algorithms fail to address: (a) noise present in bandit feedback and (b) algorithms’ heavy reliance on labeled data. To deal with the noise present in the bandit feedback, we proposed a novel algorithm named RCINE, robust to noisy bandit feedback. The methodology used in RCINE requires knowledge of the noise rates. We proposed a subroutine called NREst to estimate the noise rates, resulting in an end-to-end learning framework for learning a multiclass classifier under noisy bandit feedback. The proposed algorithm enjoys a mistake bound of the order of O( √ T) in the high noise case and of the order of O(T 2/3 ) in the worst case. We also show our approach’s effectiveness using extensive experiments on several benchmark datasets. Furthermore, to reduce the reliance of an online supervised learning algorithm on labeled data, we proposed ALBIF, an efficient stochastic sub-gradient descent algorithm for learning a multiclass classifier under an active bandit feedback setting. ALBIF enjoys a regret bound of the order O(log T) in the active learning setting as well as in the standard (non-active) bandit feedback setting. We also demonstrate the effectiveness of the proposed algorithms by conducting extensive experiments on various real-world and synthetic datasets against several benchmark algorithms.

Abstract

The rise in energy demand is a major cause of global warming and climate change. The U.S. Energy Information Administration (EIA) projects that the world energy consumption will grow by nearly 50% and the energy consumed in the building sector will increase by around 65% by 2050. There are two crucial sustainability measures, improving energy efficiency and reducing energy consumption in the building sector. To assist in this, Appliance Load Monitoring (ALM) is essential for effective load management because it determines the energy consumption and operating states of individual appliances. Energy feedback information can help reduce consumption by 5% to 15%. Compared to a single aggregate consumption feedback, a detailed individual appliance consumption feedback is more beneficial. Most electrical appliances do not have the capabilities to monitor their individual energy consumption details. To estimate this energy consumption for appliances, the two most widely used load monitoring approaches are Intrusive Load Monitoring (ILM) and Non-Intrusive Load Monitoring (NILM). ILM makes use of distributed sensing using smart power strips or power sockets to monitor individual appliances. Such kind of installation requires entering inside a house and therefore, ILM is intrusive in nature. NILM or load disaggregation is an approach to estimate individual appliance energy consumption from aggregate load measurement obtained from a single measurement point. Features extracted from electrical signals are used for automatic appliance identification. Researchers in the past have published machine learning-based ILM and NILM techniques. However, these machine learningbased ILM/NILM techniques are sensitive to noise, and their performance degrades in the presence of variations in data due to noise, sensor drift and source voltage fluctuations. The observations made by ESMI (Electrical Supply Monitoring Initiate) of Prayas Group show that grid voltages fluctuate approximately between 210 to 250 V in some regions of India. Varying voltages lead to changes in appliance power consumption patterns. Appliance identification models may not recognize appliances correctly because of their modified consumption patterns. The ILM and NILM approaches have to be tested in these varying operating conditions. There is a need to develop a low-cost, efficient and accurate load monitoring solution. This thesis investigates the robustness of supervised load identification techniques for improving ILM and NILM approaches. The first part of the thesis provides techniques to improve the ILM approach, especially to monitor plug loads. Plug loads are the appliances that are plugged in AC sockets. Plug loads account for 20% to 30% of building energy consumption, and there is an increasing trend in plug load consumption. The energy performance of buildings can be improved by effective plug load monitoring and control. A plug load monitoring field study is conducted in an academic institute to gain insights into plug load usage and energy consumption patterns. Large scale deployment of such load monitoring solutions is costly. Without load-sensing capabilities, it isn’t easy to track some plug loads that keep changing locations. Smart plug strips or smart sockets with plugged-in load identification capability are helpful for monitoring loads automatically. Smart strips or sockets record electrical measurements using an energy metering sensor. The extracted features or load signatures are used to train machine learning models for load identification. Comparative performance analysis of these load identification techniques is provided. The results show that a simple algorithm like KNN on low-frequency data of two minutes provides excellent identification accuracy. However, the identification models had problems in the presence of voltage variations. A novel Regression-based Nearest Neighbour (RBNN) Classifier is developed for robust plug load identification under varying voltages. Regression technique is used to recognize the behavior of plug loads based on their energy consumption signature and then use identified behavior for classification. The proposed algorithm is evaluated against the traditional classification algorithm on a dataset of 70 plug loads operating in varying voltage conditions. Experimental results show that the proposed algorithm performs better than standard classifiers in most of the cases. Also, a prototype smart power strip with the proposed load identification technique is designed. The remaining part of this thesis describes ways to improve event-based NILM techniques. An event-based NILM approach identifies ON or OFF events and extracts features to build appliance identification models. A study of event-based NILM algorithms under the influence of varying voltages is performed. Feature extraction and feature selection techniques are developed to reduce the impact of voltage variation. Appl

Abstract

In a blockchain network, full nodes aid in decentralization by validating incoming transactions and also by assisting new nodes who want to join the network. However, due to the high storage cost incurred by full nodes, the total number of full nodes in a system is falling. In order to curb this problem and maintain decentralization, we present Secure Raptor Encoder and Decoder (S-RED) built on Raptor Codes, which in addition to reducing storage costs, also has constant decoding time with respect to k (number of input blocks). We also present error-resilient double-peeling decoder that can recover original blocks with a high probability, even under the presence of adversaries. Blockchain popularity and application is not a new thing but blockchain is bounded by its physical capacity and is unable to accommodate the incoming traffic leading to high transaction confirmation latency. Hence a new data structure based on the direct acyclic graph is proposed popularly called blockdag which doesn’t limit blockchain physical capacity. Prism is a ledger based on blockdag technology, however, in Prism full nodes have to pay a high cost in terms of storage space due to the requirement of more blocks as compared to conventional blockchain. In order to reduce such high storage costs, we present Secure Luby Transform Raptor Encoder and Decoder (S-LTRED) built on LT and Raptor Codes. We present tri-peel decoder that can recover original blocks with a high probability, even under the presence of adversaries. S-LTRED shows to achieve a much lower bootstrap cost as compared to Prism and achieves this by reducing the storage consumption.

Abstract

We perceive the natural world through multiple input sources, with vision and language being one of the most important. The fields of computer vision and natural language processing have experienced significant progress along with the advancement of high computing power and advanced techniques in deep learning. At the same time, the intersection of the two modalities is still in its nascent stages. In this thesis, we aim to explore multimodality through several multimodal problems to understand better the shortcomings of the current state of research in the field. We also explore the multimodal problems from a lingual perspective as the current focus of majority multimodal efforts are for English only. In brief, we explore multimodal propaganda detection, multimodal machine translation, and visual question answering extensively. First, we explore a multimodal classification task in which we train models for the task of propaganda detection in memes. We also explore the importance of visual modality in the task and propose a methodology to exploit robust textual transformers using an ensemble of text and vision-language transformers to improve the performance in textually dominant tasks. Second, we explore the task of multimodal machine translation. We propose a methodology to exploit the pre-training of a textual machine translation system by bringing the visual cues to a textual domain by extracting object tags from the image. Finally, we explore the task of visual question answering in which we propose a methodology to train non-English multimodal transformers through knowledge distillation using a machine-translated dataset. We scoped the work to visual question answering, but it can also be extended to other visionlanguage tasks. To the best of our knowledge, we propose state-of-the-art models for multimodal machine translation for the English-Hindi language pair. Further, we also propose a novel approach to train visual-lingual models in different languages with the aid of machine translation and achieve state-of-the-art performance in the apanese and Hindi visual question answering tasks.

Abstract

Artificial Intelligence (AI) has seen a significant growth in the past decade in various wide ranging domains. Machine learning (ML) methods have been applied in healthcare for various tasks like disease diagnosis, disease progression and outcome prediction, etc. ML aids healthcare systems by executing medical tasks with expert-level performance in lesser time, thereby enabling optimized resource allocation, targeted care and accelerated decision-making. This thesis presents two different applications of AI in pulmonology, a domain of healthcare that deals with respiratory diseases. The first application of ML in pulmonology is about how ML is employed for early mortality prediction of COVID-19 patients. The coronavirus disease 2019 (COVID-19), caused by the virus SARS-CoV-2, is an acute respiratory disease that has been classified as a pandemic by the World Health Organization (WHO). The sudden spike in the number of infections and high mortality rates have put immense pressure on the public healthcare systems. Hence, it is crucial to identify the key factors for mortality prediction to optimize patient treatment strategy. Since routine blood test results are more widely available compared to other forms of data like X-rays, CT-scans and ultrasounds, this thesis focuses on ML methods based on blood tests data to predict COVID-19 mortality risk. A powerful combination of five features: neutrophils, lymphocytes, lactate dehydrogenase (LDH), high-sensitivity C-reactive protein (hs-CRP) and age helps to predict mortality with 96% accuracy. Various ML models (neural networks, logistic regression, XGBoost, random forests, SVM and decision trees) have been trained and their performances have been compared to determine the model that achieves consistently high accuracy across the days that span the disease. The best performing method using XGBoost feature importance and neural network classification, predicts with an accuracy of 90% as early as 16 days before the outcome. Robust testing with three cases based on days to outcome confirms the strong predictive performance and practicality of the proposed model.A detailed analysis and identification of trends was performed using these key biomarkers to provide useful insights for intuitive application. This thesis provide solutions that would help accelerate the decision-making process in healthcare systems for focused medical treatments in an accurate, early and reliable manner. In the second application of ML in pulmonology, we devise an evaluation metric called FXD score to evaluate synthetic Chest X-Ray (CXR) generation methods. In addition to the role of CXRs in diagnosing multiple disease conditions, their importance has immensely increased during the COVID-19 pandemic. Generative models are widely used for data augmentation in solving tasks with insufficient data. Evaluating generative models is inherently a challenging problem. Frechet Inception Distance (FID) and ´ other sample based evaluation metrics compare the distributions of real and generated images in the latent space of models pre-trained on the ImageNet dataset, for instance the pre-trained Inception v3 (Iv3) model. These metrics work well in analyzing the quality of synthetic images of common entities, but generally fail to analyze quality of images from niche domains. In this thesis, we propose an improved evaluation metric for evaluating the quality of synthetic CXR images generated - Frechet Distance that is calculated ´ using the pre-trained TorchXRayVision (XRV) model’s representations (FXD score). We perform a series of carefully designed experiments for evaluating the necessary conditions for a meaningful metric. Through these experiments, we show that the proposed method has more discriminability, robustness to transformations, and is more sensitive to mode dropping, mode collapsing and overfitting. Therefore, FXD score is proposed as a better alternative to metrics based on Iv3 embeddings. We demonstrate the importance of choosing domain specific representations in evaluating the quality of synthetic images by analyzing Chest X-Rays (CXRs). The FXD score will enable researchers in building generation models more efficiently which will in turn aid in analyzing infected patients more accurately

Abstract

Uber drivers in India are witnessing a slow emergence of everyday work formalization predominantly through their engagement with the Uber platform. If a vast segment of the informal employment sector in a country resembles gig work, can companies like Uber bring organizational ability and wage regularity to employment? Despite Uber’s economic model being challenged on several labor fronts in scholarship from the global North, we argue that everyday interactions with the Uber platform are ushering organized work practices and improved financial stability for drivers who formerly hailed from the informal employment sector in India. The everyday driving for Uber is filtered through a conceptual and practical work model drivers gain with due experience of Uber’s platform features. Our study uncovers the relationship between the controlling demands of the Uber platform and ensuing driver work adaptations. Drawn from a two-year study of 156 Uber drivers, in the metropolis of Mumbai, Chennai, Delhi, and Hyderabad, the thesis examines platform-driver interactions, outcomes for every day, and resulting structural shifts in employment for the Uber driver in the face of challenges in the Indian market. Our main argument focuses on Uber’s ride-hailing platform bringing a certain degree of formalization to a driver’s everyday work. We present our findings from a qualitative investigation of Uber ride-hailing services impacting drivers to 1. optimize earnings 2. link work effort to wages 3. converge towards platform compliance. We posit engaged and persistent interactions with the Uber platform bring structure and formality to the profession of driving for Indian Uber drivers. Finally, our study aims to balance an all-encompassing critique of platform-driven economy, mainly arising from scholarship in the global North, with a social and contextually driven study of Indian Uber drivers and their approaches to ridehailing service work. The thesis primarily offers an emic view of driver participation and engagement with the Uber platform, their everyday work structures, and employment conditions.

Abstract

In linguistics, discourse is a unit of language longer than a single sentence, referring to spoken or written language in social contexts. Critical Discourse Analysis (CDA) is an interdisciplinary research area involving linguistics and social sciences which aims to explore the links between language use and social practice. It is a qualitative approach used to describe, interpret, and explain how discourses construct and maintain social interactions. In this thesis, we introduce the concepts of CDA and apply them to focus mainly on the analysis of Event and Political Discourses. We also explore the concept of Political Bias in news and attempt to have a better understanding of it by identifying, categorising and quantifying political bias in news articles. Throughout the thesis, we detail the process of curating and developing relevant datasets wherever necessary. In the Event Discourse Analysis, we develop a dataset, Manovaad, which consists of events belonging to domains like sports, politics, entertainment, trade, etc., and their news coverage at different levels of reporting. Using a Bidirectional-Recurrent Neural Network (Bi-RNN), we calculate the subjectivity of each news article and then understand trends in the variation of subjectivity across local, national, and international news reports for a given event. We also analyse how the focus of the news report related to an event shifts as time passes, and compare the trends of focus shift across different levels. This analysis about focus shifts and subjectivity in reporting helps understand how significance associated with an event in terms of features like focus, perspective, inclination, and details is dependent largely on the closeness of the event with the person or the organisation reporting it. In Political Discourse Analysis, we begin with a comparative case study of the election speeches given by the winning and losing party leaders in the 2014 Andhra Pradesh elections. We compare and contrast their interpersonal speech choices, highlight interesting patterns at both word and sentence levels, and how they affect the audience. The next part of our work on Political Discourse deals with Political Bias in the news. We often observe that in newspapers and news websites, the reporters tend to emphasize more on particular viewpoints selectively and present biased information aligned with their personal political ideology. This can lead to widespread alteration of mass political opinion and impact the decision of the voters. Our work on political bias is two-part. Firstly, we propose a method to enhance the performance of the current best bias detection network by incorporating linguistic information in the form of presupposition, which significantly outperforms the state-of-the-art for elugu.However, Political news is a domain where the majority of the news tends to have some bias. As a result, just detection of bias is not sufficient. For a reader to have a complete understanding of bias in a news article, it is essential to have insights into the ways in which bias is propagated, and an idea about the magnitude of bias. This forms the next part of our work, where we design a categorisation schema for Political Bias and build an annotated public corpus - PoBiCo-21. Along with it, we propose a novel Sentiment-Based Ranking Mechanism for the quantification of Political bias. The majority of the work in this thesis has been done on Telugu language data. We highlight the challenges of working with a low-resource language and demonstrate how we can combine linguistic and computational techniques such as Machine Learning and Deep Learning to achieve the desired performance.

Abstract

The insulin hormone stored in hexameric form, and dissociates to dimeric form and finally to the active monomeric form. When insulin secretion is impaired whereby affecting various metabolic processes, as a final response insulin analogs are subcutaneously injected before meals to facilitate glucose metabolism. Depending on the molecular details, analogs are rapid or slow acting based on the dissociation rate of dimer to monomer. Insulin Aspart is a recombinant human insulin analog, acting faster than regular human insulin. Despite its practical and elementary importance, the process of insulin aspart dimer dissociation is relatively unknown. Here, we combined molecular dynamics simulations and umbrella sampling to characterize the energetic and structural features of dissociation of insulin aspart dimer. Like previous studies on human insulin (another well studied analog), insulin aspart can also display wide spectrum of pathways for dimer dissociation: from dissociation happening without major change in monomer structure to dissociation that is coupled with unfolding of protein. Additionally, water plays a vital role in the dissociation of the insulin aspart by stabilizing the monomers in the dissociated state. Our study shows the molecular details, such as the variation in the structural and orientation and conformational changes along the minimum energy pathways in the process of dissociation of insulin aspart dimer.

Abstract

Code-mixing refers to the mixing of two or more languages where words from different languages are interleaved with each other in the same conversation. Code-mixing is popular in multilingual societies around the world and is commonly used in social media texts. In recent times, with the increasing popularity of social media platforms, there has been an increase in the usage of code-mixed languages. English-Hindi code-mixed language colloquially called Hinglish is very commonly used in India because of a large number of bilingual speakers who use English in their professional lives and Hindi in their personal lives. Traditional Natural language processing systems, which are usually trained on monolingual corpora, do not perform well with code-mixed texts. Building NLP systems for code-mixed languages could help build interactive systems for the large number of people using code-mixed languages. It would also enable us to build systems that can process the large amount of code-mixed data generated on social media and make the content accessible for people and machines alike. For example, building translation systems for code-mixed languages would allow people to communicate with each other comfortably. Other tasks such as sentiment analysis or toxic content detection can enable high-level analysis of code-mixed texts. In this thesis, we propose a simple approach for machine translation of Hinglish (English Hindi code-mixed) texts to English and explore how transfer learning from Hindi to English translation task can improve the performance of our translation system. We also propose an approach for translating English data to Hinglish. We provide parallel Hindi translations along with the English sentences as input to our translation system and analyse the improvement in the performance of our system. We also propose BLEUnormalized - a modified version of the BLEU metric to evaluate Hinglish outputs which often have informal transliterations and varying degrees of code-mixing. Further, we propose a method for sequence-level classification tasks for Hinglish data using our Hinglish to English translation system. First, we translate the Hinglish data to English, and then we use transfer learning from state-of-the-art English models for processing the translated data. We evaluate the effectiveness of our approach on the tasks of Sentiment Analysis and Natural Language Inference of Hinglish texts. We use various pre-trained models that have been fine-tuned for similar English-only tasks and have shown state-of-the-art performance in those. We further fine-tune these models on the translated datasets. To the best of our knowledge, we achieve state-of-the-art performance in both tasks. Finally, we explore Natural Language Inference in a different setting, called tabular entailment, where the task is to verify a statement given a table as evidence. We propose a method for classifying whether the table supports the statement based on TAPAS. We evaluate how transfer learning and standardizing tables to have a single header row can improve TAPAS’ performance. We also propose a method for predicting which cells in the table provide evidence for/against the statement. We again use TAPAS and evaluate how different fine-tuning strategies can improve TAPAS’ performance

Abstract

Cancer has been the leading cause of death across the globe and there has been a constant push and research from the scientific community as a whole towards assisting its diagnosis. Particularly the last decade has seen widespread use of computer vision and AI towards tackling cancer diagnosis using both radiological (non-invasive, ex: X-ray, CT ) and pathological modalities (invasive ex: histopathology). A histopathologic Whole Slide Image (WSI) represents digitized image of tissue sample characterized by a large size of up to 109 pixels at maximum resolution. It is considered the gold standard for cancer diagnosis. The routine diagnosis involves experts called pathologists to analyse a slide containing the tissue samples under a microscope. The process is often subjective to cognitive load and the diagnosis is prone to inter and intra pathologist errors. With the digitization of tissue samples as WSI, computerassisted diagnosis would be impactful to address the above issues especially with the advent of deep learning. For this we require models to be trained with large amounts of annotated data as well as understanding the cancer manifestation across organs. In this thesis work, we address two major issues with the help of deep learning techniques. (1) assisting the whole slide image annotation using an expert in the loop and (2) understanding the relationship between cancers and bring to light commonalities of cancer patterns between certain pairs of organs. A typical process of slide diagnosis under a microscope by experts involves exhaustive analysis in scanning across the slides, searching for anomalous/tumorous regions present. Owing to the large dimensions of the histopathology WSI, visually searching for clinically significant regions (patches) is a tedious task for a medical expert. Sequential analysis of several such images further increases the workload resulting in poor diagnosis. A major impediment to automate this task using a deep learning model is the requirement of large amounts of annotated data of WSI patches which is a laborious process and involves exhaustive search for anomalous regions. To tackle this issue, the first part of the thesis work proposes a novel CNN-based, expert feedback-driven interactive learning technique to mitigate this issue. The proposed method seeks to acquire labels of the most informative patches in small increments with multiple feedback rounds to maximize the throughput. It requires the expert to query a patch of interest from a slide and provide feedback to a set of unlabelled patches chosen using the proposed sampling strategy from a ranked list. The proposed technique is applied in a setting that assumes there exists a large cohort of unannotated slides, almost eliminating the need of annotated data upfront; instead learns with an expert involvement. We discuss several strategies for sampling the right set of samples to be labelled by the expert to minimise the expert feedback and maximise the throughput. Theproposed technique can also annotate multiple slides parallelly using a single slide under review (used to query anomalous patches), which in turn reduces the effort of annotation. The Cancer Genome Atlas (TCGA) contains large repositories of histopathology whole slide images spanning across several organs and subtypes. However, not much work has gone into analysis of all the organs and subtypes and their similarities. Our work attempts to bridge this gap by training deep learning models to classify cancer vs normal patches for 11 subtypes spanning 7 organs (9792 tissue slides) to achieve near-perfect classification performance. We used these models to investigate their performances in the test set of other organs (cross organ inference). We found that every model had a good cross organ inference accuracy when tested on breast, colorectal and liver cancers. Further, a high accuracy is observed between models trained on the cancer subtypes originating from same organ (kidney and lung). We also validated these performances by showing the separability of cancer and normal samples in a high dimensional feature space. We further hypothesized that the high cross organ inferences are due to shared tumor morphologies among organs. We validated the hypothesis by showing the overlap in the Gradient-weighted Class Activation Mapping (GradCAM) visualizations and similarities in the distributions of nuclei geometrical features present within the high attention regions.