- Privacy-Preserving Technologies in Data
- Recommender Systems and Techniques
- Complex Network Analysis Techniques
- Ethics and Social Impacts of AI
- Explainable Artificial Intelligence (XAI)
- Bayesian Modeling and Causal Inference
- Domain Adaptation and Few-Shot Learning
- Topic Modeling
- Advanced Causal Inference Techniques
- Opinion Dynamics and Social Influence
- Advanced Bandit Algorithms Research
- Oil and Gas Production Techniques
- Adversarial Robustness in Machine Learning
- Digital Marketing and Social Media
- Hydraulic Fracturing and Reservoir Analysis
- Privacy, Security, and Data Protection
- Machine Learning and Data Classification
- Reservoir Engineering and Simulation Methods
- Digital Mental Health Interventions
- Technology Adoption and User Behaviour
- Social Media and Politics
- Data Stream Mining Techniques
- Text and Document Classification Technologies
- Spam and Phishing Detection
- IoT and Edge/Fog Computing
Lovely Professional University
2020-2024
Birla Institute of Technology and Science, Pilani
2023-2024
Nvidia (United States)
2024
Institute of Management Technology
2023-2024
Swiss Federal Laboratories for Materials Science and Technology
2023-2024
Sawai ManSingh Medical College and Hospital
2022-2024
Teerthanker Mahaveer University
2023-2024
Chaudhary Charan Singh University
2024
Gyan Vihar University
2023
Microsoft Research (India)
2022-2023
There is an exponential growth in textual content generation every day today's world. In-app messaging such as Telegram and WhatsApp, social media websites Instagram Facebook, e-commerce like Amazon, Google searches, news publishing websites, a variety of additional sources are the possible suppliers. Every instant, all these produce massive amounts text data. The interpretation data can help business owners analyze outlook their product, brand, or service take necessary steps. development...
The causal capabilities of large language models (LLMs) is a matter significant debate, with critical implications for the use LLMs in societally impactful domains such as medicine, science, law, and policy. We further our understanding their implications, considering distinctions between different types reasoning tasks, well entangled threats construct measurement validity. LLM-based methods establish new state-of-the-art accuracies on multiple benchmarks. Algorithms based GPT-3.5 4...
Recommender systems associated with social networks often use explanations (e.g. "X, Y and 2 friends like this") to support the recommendations. We present a study of effects these in music recommendation context. start an experiment 237 users, which we show varying levels information analyze their effect on users' decisions. distinguish between two key decisions: likelihood checking out recommended artist, actual rating artist based listening several songs. find that while do have some...
How predictable is success in complex social systems? In spite of a recent profusion prediction studies that exploit online and information network data, this question remains unanswered, part because it has not been adequately specified. paper we attempt to clarify the by presenting simple stylized model attributes error one two generic sources: insufficiency available data and/or models on hand; inherent unpredictability systems other. We then use motivate an illustrative empirical study...
To construct interpretable explanations that are consistent with the original ML model, counterfactual examples---showing how model's output changes small perturbations to input---have been proposed. This paper extends work in by addressing challenge of feasibility such examples. For models critical domains as healthcare and finance, examples useful for an end-user only extent perturbation feature inputs is feasible real world. We formulate problem preserving causal relationships among input...
Recommendation systems are an increasingly prominent part of the web, accounting for up to a third all traffic on several world's most popular sites. Nevertheless, little is known about how much activity such actually cause over and above that would have occurred via other means (e.g., search) if recommendations were absent. Although ideal way estimate causal impact randomized experiments, experiments costly may inconvenience users. In this paper, therefore, we present method estimating...
In the domain generalization literature, a common objective is to learn representations independent of after conditioning on class label. We show that this not sufficient: there exist counter-examples where model fails generalize unseen domains even satisfying class-conditional invariance. formalize observation through structural causal and importance modeling within-class variations for generalization. Specifically, classes contain objects characterize specific features, can be interpreted...
In addition to efficient statistical estimators of a treatment's effect, successful application causal inference requires specifying assumptions about the mechanisms underlying observed data and testing whether they are valid, what extent. However, most libraries for focus only on task providing powerful estimators. We describe DoWhy, an open-source Python library that is built with as its first-class citizens, based formal framework graphs specify test assumptions. DoWhy presents API four...
Many technologies, such as surface-acoustic-wave (SAW) resonators, sensors, and piezoelectric MEMS require highly-oriented textured functional thin films. The best results are typically achieved for on-axis sputter geometries, but in some scenarios, this is not feasible, during co-deposition from multiple magnetrons or when coating substrates with high aspect ratios. Ionized physical vapor deposition (PVD) techniques HiPIMS can be used to accelerate the film-forming species onto growing film...
This paper describes an Electrooculogram (EOG) and gaze based hands-free natural interaction system design for virtual reality (VR) games which enhances the immersive VR experience. The traditional interfaces like joysticks, mouse, keyboards, hand-worn data-gloves when used with HMD peripherals are obtrusive is a step further towards building see-and-play user in games. interface provides enhanced gaming experience user's environment interacting as per eye movements. online blink detection...
Many online services, such as search engines, social media platforms, and digital marketplaces, are advertised being available to any user, regardless of their age, gender, or other demographic factors. However, there growing concerns that these services may systematically underserve some groups users. In this paper, we present a framework for internally auditing differences in user satisfaction across groups, using engines case study. We first explain the pitfalls naively comparing...
Mental illness is a global health problem, but access to mental healthcare resources remain poor worldwide. Online peer-to-peer support platforms attempt alleviate this fundamental gap by enabling those who struggle with provide and receive social from their peers. However, successful requires users engage each other failures may have serious consequences for in need. Our understanding of engagement patterns on limited critical inform the role, limitations, design these platforms. Here, we...
Can we predict the future popularity of a song, movie or tweet? Recent work suggests that although it may be hard to an item's when is first introduced, peeking into its early adopters and properties their social network makes problem easier. We test robustness such claims by using data from networks spanning music, books, photos, URLs. find stronger result: not only do predictive models with achieve high accuracy on all datasets, they also generalize well, so much trained any one dataset...
As per recent progress, online social network (OSN) users have grown tremendously worldwide, especially in the wake of COVID-19 pandemic. Today, OSNs become a core part many people’s daily lifestyles. Therefore, increasing dependency on encourages privacy requirements to protect from malicious sources. contain sensitive information about each end user that intruders may try leak for commercial or non-commercial purposes. ensuring different levels is vital requirement OSNs. Various...
<div class="section abstract"><div class="htmlview paragraph">This paper introduces an innovative digital solution for the categorization and analysis of fractures in Auto components, leveraging Artificial Intelligence Machine Learning (AI/ML) technologies. The proposed system automates fracture process, enhancing speed, reliability, accessibility users with varying levels expertise.</div><div paragraph">The platform enables to upload images fractured parts, which are...
Decentralized wireless networks with self-configuring nodes that can interact without the need for fixed infrastructure are known as mobile ad hoc networks, or MANETs. Reliable and effective communication is a difficulty traditional routing methods because of node mobility, bandwidth limitations, energy limits. A potential remedy these issues use machine learning (ML) artificial intelligence (AI) into design protocols. MANETs optimize choices, adjust flexibly to shifting network conditions,...