ABSTRACT. Artificial Intelligence has been widely used during the last two decades and has remained a highly-researched topic, especially for complex real-world problems. Evolutionary Intelligence (EI) techniques are a subset of artificial intelligence, but they are slightly different from the classical methods in the sense that the intelligence of EI comes from biological systems or nature in general. The efficiency of EI is due to their significant ability to imitate the best features of nature which have evolved by natural selection over millions of years. The central theme of this presentation is about EI techniques and their application to complex real-world problems. On this basis, first I will talk about an automated learning approach called genetic programming. Applied evolutionary learning will be presented, and then their new advances will be mentioned. Here, some of my studies on big data analytics and modelling using EI and genetic programming, in particular, will be presented. Second, EI will be presented including key applications in the optimization of complex and nonlinear systems. It will also be explained how such algorithms have been adopted to engineering problems and how their advantages over the classical optimization problems are used in action. Optimization results of large-scale towers and many-objective problems will be presented which show the applicability of EI. Finally, heuristics will be explained which are adaptable with EI and they can significantly improve the optimization results.

ABSTRACT. With the astounding growth in the artificial intelligence (AI) space, a variety of automation and acceleration problems across industry verticals are being attempted and addressed. Today we have a bevy of pioneering AI algorithms and models. With the ready availability of big data and greater computational power, AI-based data analytics to emit predictive, prescriptive and personalized insights in time becomes the new normal.

There are state-of-the-art platforms for AI model engineering, evaluation, optimisation and deployment across cloud environments. Now with the exponential growth of connected devices (alternatively referred to as networked embedded systems or IoT edge devices) joining mainstream computing in the digital era, the computing activity is being systematically shifted to IoT edge devices. Edge devices individually and collectively perform proximate data processing to extract timely and actionable insights, which, in turn, results in a slew of real-time and real-world services and applications. By translating heavyweight AI models into lightweight models using a suite of compression techniques and tools, hosting and running AI models on edge devices and their clusters become the talk of the town. Such a transition empowers edge devices to be intelligent in their operations, offerings and outputs. In this talk, I would like to demystify the edge AI paradigm and how it is going to be a game-changing phenomenon for the entire society. Further on, I will focus on detailing some prominent personal and professional use cases of edge AI. Especially setting and sustaining intelligent environments and enterprises is being simplified and speeded up through the smart leverage of the distinct power of the edge AI paradigm

ABSTRACT. This talk introduces a novel approach to GPTFX, an AI-based mental detection with GPT frameworks. This approach leverages GPT embeddings and the fine-tuning of GPT-3. This approach exhibits superior performance in both classifying mental health disorders and generating explanations with an accuracy of around 87% in classification and Rouge-L of around 0.75. We utilized GPT embeddings with machine learning models for the classification of mental health disorders. Additionally, GPT-3 was fine-tuned for generating explanations related to the predictions made by these machine learning models. Notably, the proposed algorithm proves well-suited for real-time monitoring of mental health by deploying AI-IoMT devices, as it has demonstrated greater reliability when compared to traditional algorithms.

ABSTRACT. Pothole repair stands as a critical component of road maintenance, and the ongoing challenge faced by road management authorities involves the continuous monitoring of road surfaces. Currently, the detection of potholes relies on labour-intensive and time-consuming manual image/data processing. Furthermore, as traffic continues to grow exponentially, roads are becoming increasingly susceptible to damage within shorter time frames. However, there is an opportunity to address this challenge by leveraging computer vision, which can automate the visual inspection process using digital imaging to identify potholes from a sequence of images. The persistent issue of potholes and poor road quality threatens road safety and infrastructure integrity. This issue is underscored by alarming accident statistics, with thousands of fatalities and injuries reported annually. In this paper, a visual sensing pothole detection model, built upon YOLO V8, has been proposed. Additionally, road quality and PHR mapping via sensory data (accelerometer and gyrometer), which undergoes post-processing to provide quality observations for various vehicles and classify them, has been introduced. Leveraging this data, concerned officials were notified via the developed backend service at certain intervals about the quality and repairs required at specific locations. This streamlines the government’s efforts by deducing survey time, cost, and resource expenditure since the data is collected by vehicles on the road. Additionally, it notifies users of any upcoming potholes and road conditions ahead.

ABSTRACT. Medical Question Answering is a significant under- taking, as it could result in more rapid and precise diagnoses and treatment decisions. Question Answering models have made significant advances in accurately answering medical questions and supporting healthcare professionals in finding relevant in- formation from vast amounts of medical literature and patient records. This paper presents a critical evaluation of the leading- edge technology in question answering using natural language processing in the field of medical. The review covers a different technology related to QA in the medical domain, including datasets, models, evaluation metrics, and applications. It also covers the role of large language models (LLMs) like ChatGPT in medical question answering. Furthermore, the review dissects the different metrics and methodologies that are used to assess the performance of medical QA systems, with a particular focus on metrics that capture the medical relevance and accuracy of the answers. Finally, the review explores the future directions and research challenges in medical QA using NLP. Overall, This review comprehensively summarizes the state-of-the-art in ques- tion answering using NLP in the medical domain, highlighting the accomplishments, challenges, and potential future advances. It is a valuable and powerful resource for researchers, practitioners, and healthcare professionals who are curious about using NLP techniques to improve healthcare delivery by answering medical questions.

ABSTRACT. In federated learning (FL), the idea is to train and bring out a single global model collaboratively with the aid of numerous client machines and devices while everything is being coordinated by a central server. However, given the variability of the data, developing a single global model could be problematic for some clients taking part in federated learning. Therefore, in order to deal with the difficulties brought in by statistical heterogeneity and the non-Informally, Identically Dis- tributed (IID) distribution of data, the personalization of the global model becomes essential. In contrast to the earlier research works, we suggest a novel method for creating a customized model. This further encourages all clients to take part in federation even in the presence of statistical heterogeneity. Such an arrangement is to enhance the performance as opposed to serving just a resource for the central server’s model training. In order to achieve this personalization, we use hybrid pruning which is a combination of structured and unstructured pruning to identify a small subnetwork for each client. Each pruning technique has been implemented based on the sparsity %. In this proposed work, we have shown the experimental implementation of pruning techniques and their evaluation to reduce the communication cost. This work will also help FL process to work on low bandwidth of the Internet connection.

ABSTRACT. The field of multimodal sentiment analysis is experiencing significant growth as it extends to integrating natural language processing, computer vision, and emotional computing. This initiative aims to facilitate machines’ capacity to comprehend and analyze emotions conveyed through diverse modalities, encompassing written language, spoken communication, visual representations, and audiovisual content. Incorporating various modalities facilitates a more extensive comprehension of emotions, enhancing interactions between humans and computers through empathy and context awareness. This paper explores current advancements in multimodal sentiment analysis, examining the difficulties encountered, the methodologies employed, and the various applications within this domain. This study explores cross-modal fusion techniques, contextual analysis, and strategies for addressing data sparsity and missing modalities. The advancement of multimodal sentiment analysis has resulted in its widespread influence across various domains. These domains include customer service, social media analysis, healthcare, and education. Ultimately, this progress contributes to the development of emotionally intelligent interactions between humans and machines in a world that is becoming increasingly interconnected. This paper explores some challenges and future research areas in multimodal sentiment analysis.

ABSTRACT. In this paper a prototype model for creating a smart world based on a number of quantum data processing technologies has been presented. It also attempts to define the roles of human beings in this novel world highlighting the various job roles which would be required. It also attempts to define several avenues where reskilling is required to survive in such an environment. The Quantum Smart World Prototype is 45% more efficient compared to the conventional smart world prototype.

ABSTRACT. The Machine Learning and Deep Learning Models require a lot of data for the training process, and in some scenarios, there might be some sensitive data, such as customer information in-volved, which the organizations might be hesitant to outsource for model building. Some of the privacy-preserving techniques such as Differential Privacy, Homomorphic Encryption, and Secure Multi-Party Computation can be integrated with different Machine Learning and Deep Learning algorithms to provide security to the data as well as the model. In this paper, we pro-pose a Chaotic Extreme Learning Machine and its encrypted form using Fully Homomorphic Encryption where the weights and biases are generated using a logistic map instead of uniform distribution. Our proposed method has performed either better or similar to the Traditional Extreme Learning Machine on most of the datasets.

ABSTRACT. To improve network management in 5G networks by utilizing network slicing and traffic classification, we propose explainable machine learning techniques. For achieving interpretability, we devised a novel, general-purpose filter-based feature selection method, namely, extended t-statistic suitable for multi-class classification problems. Real-world network traffic is used to test the effectiveness of our approach. We found that our approach achieved high accuracy while also providing interpretability. This is significant because our approach can improve the overall performance of 5G networks by enhancing the accuracy and trustworthiness of network slices and traffic classification.

ABSTRACT. Measurement data acquired from an industrial heat exchanger (HE) can provide a reliable inference on its dynamics and performance. Harnessing the information available on the acquired data, the data-driven modeling technique can create a dynamic model of an HE without any expert knowledge or complex mathematical analysis. Motivated by these factors, data-driven modeling techniques are investigated to model an industrial Naphtha cooler HE in this work. Model describing the HE dynamics is a primary requirement to simulate control strategies and enhance operational efficiency. Naphtha cooler is an industrial shell and tube HE, where the temperature of Naphtha available at the shell side is controlled using the cooling water flow rate at the tube side. Hence, the HE is modeled as a Multi-input Multi-output (MIMO) process with four inputs and two outputs. Three different data-driven modeling approaches namely, traditional, polynomial, and non-linear are investigated. A total of eight model structures are considered and the fit percentage is used as performance metrics to determine the accurate model. Experimental results illustrate that the ARMAX model exhibits a significant 88.2% fit on average.

ABSTRACT. Significant progress has been achieved, in the field of marine science, which entails the complex study of the enormous ocean and its interrelated ecosystem through the incorporation of data science. This study focuses on the application of data science in marine science, particularly in predicting anomalies in sea surface temperature (SST). Accurately predicting SST anomalies is critical for understanding climate dynamics, oceanic currents, and the delicate balance of marine life. This study investigates current advances in SST anomaly prediction with data science through a meticulous analysis of research articles. It presents a thorough examination of machine learning techniques for forecasting SST anomalies, highlighting their advantages, disadvantages, and potential future directions. This study adds to our understanding of ocean behavior by fusing cutting-edge analytical techniques with a deep knowledge of SST anomalies. By critically analyzing previous studies and offering novel insights, this research contributes to advancing the field of marine data science and supports the sustainable management of marine resources. In this research endeavor, we introduce an innovative ensemble framework that amalgamates a meticulously designed 9-layer Long Short Term Memory (LSTM) network with the robust eXtreme Gradient Boosting (XGBoost) ensemble algorithm. By synergizing these two advanced methodologies, we aspire to propel the realm of marine data science towards greater horizons. Our approach distinctly contributes to the domain by substantially refining the prediction of SST anomalies, consequently culminating in a noteworthy enhancement in the precision of forecasts.

ABSTRACT. In this research work we have examined the top 5 NIRF ranked Government and private universities along with 5 worldwide prestigious universities. Our study is focused on analyzing some of the SEO Parameter such as, Referring Domains, Organic and Paid search, Backlinks Analysis, Keyword and Backlink gap. These are some of the crucial factors that contributes the develop-ment of SEO and enhancement of website ranking in the search engine result page (SERP). In this research we have used some of the online SEO Tools pre-sent in the search engines such as SEMrush Similar Web and Uber Suggest. It examines the websites of the top worldwide prestigious universities. From our findings we found that “IIT Roorkee” lead 50.92% in organic search and 2.53% lead in referral traffic search compared to other top ranking IIT’s, In private uni-versities “Vellore Institute of Technology” (VIT) receives 3.725 million users as Monthly visits and in global universities Imperial College leads 1.46% in Paid search and “Massachusetts Institute of Technology” (MIT) leads with 4.13% in social media traffic. These are some of the outcomes of our findings. In this paper we will be analyzing the various factors of SEO on top national and international universities website using SEO tools.

ABSTRACT. The adoption of Multi-User Multiple Input Multiple Output (MUMIMO) tech in IEEE 802.11 standards transformed wireless communication, enabling data transmission to multiple users simultaneously. However, traditional antennas struggle to meet MU-MIMO's precise beamforming needs. Additionally, the environmental impact of non-biodegradable antennas adds complexity. This research introduces an eco-friendly antenna design using biodegradable plant fibers like cotton, hemp, bamboo, and jute. This sustainable antenna not only addresses e-waste concerns but also offers a long-term, eco-conscious solution for precise beamforming. Experimental analysis validates its effectiveness, showcasing potential to revolutionize wireless tech with minimal environmental impact

ABSTRACT. The IoMT is a device that works with different health care systems to help develop technologies. It reduces the unnecessary burden of visiting the hospital. Instead it enables the transfer of medical-related data with a physician through a secured network. The Smart devices, such as wearable and medical monitors are used for health care use in the IoMT network. Telehealth and other services are adapted in homes and hospitals in other real-time locations. Approximately the globe, 80% of health care have implemented IoT (Internet of Things) technologies, and 20% are expected to do the same in the future. This chapter focuses on four topics: 1] recent advances on the internet of medical things,2] highlighting the emerging data fusion process for data preparation,3] discussing the life cycle of sensor devices and their protocols, and 4] processing data with different techniques and theories. In a nutshell, this chapter gives a comprehensive overview of IoMT research advancement, application, data management, and analytics techniques along with the current research challenges.

ABSTRACT. Maternal mortality rates in India are high, especially in states that belong to the empowered action group (EAG). The study focuses on identifying inter- and intra-state differences while discussing trends and patterns in maternal mortality reduction in India. Our research reveals that the developments in the maternal mortality ratio (MMR) over the past 20 years—particularly the rate of decline—have not been consistent with the apparent advancements in the nation's socioeconomic indices. Massive MMR disparities between and within states are a significant policy concern. The MMR reported for the EAG/Assam group, for example, was 438 in 2001–03 and 148 in 2017–19, over five times greater than Kerala’s (MMR 30), the state with the lowest MMR of all. High maternal mortality in India, especially in states with an empowered action group (EAG), is a serious policy problem. This study explores the patterns and trends in the decline of maternal mortality in India and emphasizes the differences between the states. It has been discovered that the developments in the maternal mortality ratio (MMR) over the previous two decades, notably the pace of fall, do not correspond well with the apparent advancements in the nation's socioeconomic metrics. A significant policy problem is the enormous in the inter - state disparities in the Maternal mortality ratio reduction.

ABSTRACT. Inspired by the principles of electronics, Moore's laws reflect the exponential growth seen in various human activities. The vast expansion of knowledge, particularly in technical fields at the higher undergraduate education level, poses challenges, contributing to increased dropout rates. This trend, influenced in part by the mindset of Generation Z, aligns with the mass and elite education division outlined in the Bologna Declaration. Dropout rates predominantly impact mass education, specifically at the undergraduate level. To address this, enhancing the effectiveness of technical higher education is crucial. A key initiative involves shifting from the traditional bottom-up approach to the top-down Systems Engineering approach. The paper suggests practical steps