Real-time data streaming pipelines are immensely valuable in today's data-driven world since they enable continuous data processing and analytics. This research paper provides a comprehensive exploration of the architecture, development, and deployment of an advanced real-time data streaming pipeline. It utilizes Docker for containerization, Apache Kafka for distributed streaming, Apache Spark for dynamic data transformation, and Cassandra for efficient NoSQL storage. The study outlines the intricacies of integrating these technologies, examining the pipeline's components, functionalities, performance metrics, and potential applications. Through this case study, the paper showcases the efficacy of open-source tools in constructing highly scalable and resilient data streaming pipelines.

Every now and then, we witness significant improvements in the performance of Deep Learning models. A typical cycle of improvement involves enhanced accuracy followed by reduced computing time. As algorithms get better at their job, it is worthwhile to try to evaluate their performance on problems that are affected by them. Computationally intense problems, such as object detection for Computer Aided Laparoscopy (CAL), can benefit from such improvements in such technologies. Recently a new set of variants of You Look Only Once (YOLO) models based on Neural Architecture Search (NAS) technique have been released. Deci, the enterprise behind this new development, touts a much better performance both in terms of accuracy as well as computational efficiency. In this paper, we have analyzed the performance YOLO-NAS on a well-known benchmark dataset related to CAL. We found that the performance of all the NAS-based YOLO was inferior as compared to other State-of-the-Art (SoTA) YOLO models. We compare our results against the YOLOv7 model too.

This article explores the transformative role of AI-powered Virtual Health Assistants (VHAs) in healthcare. VHAs, including chatbots and voice-activated systems, offer 24/7 accessible support. The piece highlights their applications in personalized online consultations, clinical decision-making, image diagnosis, and automation of repetitive tasks. Emphasizing VHAs as health coaches, the article discusses their potential to empower patients, especially those managing long-term illnesses. It foresees a promising future for virtual health, emphasizing the need to address legal and security considerations for comprehensive integration into healthcare practices.

In recent years, Sound AI is being increasingly used to predict machine failures. By attaching a microphone to the machine of interest, one can get real time data on machine behavior from the field. Traditionally, Convolutional Neural Net (CNN) architectures have been used to analyze spectrogram images generated from the sounds captured and predict if the machine is functioning as expected. CNN architectures seem to work well empirically even though they have biases like locality and parameter sharing which may not be completely relevant for spectrogram analysis. With the successful application of transformer-based models in the field of image processing starting with Vision Transformer (ViT) in 2020, there has been significant interest in leveraging these in the field of Sound AI. Since transformer-based architectures have significantly lower inductive biases, they are expected to perform better than CNNs at spectrogram analysis given enough data. This paper demonstrates the effectiveness of transformer-driven architectures in analyzing Sound data and compares the embeddings they generate with CNNs on the specific task of machine fault detection.

A document by BigDD. Click on the document to view its contents.

This paper introduces GPT-Neo-CRV, a novel adaptation of the GPT-Neo 1.5B model, incorporating a Cross-Referential Validation (CRV) module to significantly enhance the accuracy and reliability of information generated by Large Language Models (LLMs). GPT-Neo-CRV addresses the critical challenge of misinformation in LLM outputs, a growing concern in fields where precision and reliability of information are crucial. Through rigorous testing against the BIG-bench categories, GPT-Neo-CRV demonstrated marked improvements in tasks requiring factual correctness and complex reasoning, surpassing the standard GPT-Neo model. This study delves into the implications of these advancements, potential limitations, and the ethical considerations inherent in integrating factual validation mechanisms into LLMs. It highlights the need for comprehensive, unbiased, and ethically curated validation sources and emphasizes the importance of ongoing research in enhancing LLMs' adaptability, scalability, and ethical integrity. The development of GPT-Neo-CRV represents a significant step forward in the AI field, contributing to a more informed and truthful digital landscape and setting new standards for future LLM developments.

The nascent tokenized assets industry needs to develop standards around asset definition schemas that permit each industry or vertical to create profiles based on the schemas that are specific for addressing the needs of that industry. Machine-readable asset schemas and profiles are a core element of a proper tokenized assets management lifecycle. As metadata artifacts supporting tokenization, they must be managed as part of the lifecycle. A global network of decentralized registries is needed to provide persistence and accessibility to the published asset schemas and profiles. The notion of schemas, profiles, and registries aligns with the asset-referenced tokens (ART) in the recent EU MiCA Regulation on markets in crypto-assets.

Large Language Models (LLMs) have shown their capabilities in solving tasks across various domains. Recently, they have found applications in automating the design of digital circuits. However, there remains a dearth of research investigating the use of LLMs in automating the design of analog circuits. In this work, we propose LADAC: Large Language Model-driven Auto-Designer for Analog Circuits, to assist analog circuit design. LADAC designates an LLM as a decision-making agent, which formulates optimized design strategies based on user specifications and foundational principles of analog circuit design expertise. We applied LADAC to design two kinds of amplifiers with open-loop gain over 80dB, and a ring oscillator with 100MHz oscillation frequency successfully, showing that LADAC is capable of designing analog circuits. Furthermore, it demonstrates the LLMs' potential for designing a broader range of analog circuits.

The current large language models (LLMs) mainly lacks this capability: answering  high-precision questions/prompts. LLMs is actually a powerful fuzzy memory system that makes it difficult to answer high-precision questions. The results of code execution is a kind of high-precision answer. And expert system applications need to answer these kinds of questions. In this paper, to solve the above problem, We propose a design of LLMs combined with the text2code approach.

Designing Brain Computer Interfaces (BCIs), for helping patients, needs appropriate datasets which are relevant for the language of the patients. There exists a significant shortage of datasets for Indian languages that can be used for BCI research. Malayalam is a prominent south Indian language spoken by more than 34 million people, yet, there exist no BCI datasets for research. We address this issue by creating a dataset for selected Malayalam words by collecting Electro Encephalograph (EEG) signal samples. Our dataset was created by generating EEG samples using the OpenBCI Cyton device when the commonly used Malayalam words were spoken by a volunteer. The created dataset consists of three major types of data: (i) EEG data for spoken Malayalam words, (ii) EEG data for the spoken English words which were closest to the English translation of the corresponding Malayalam words, and (iii) EEG data for sub-vocal (silent) pronunciation of the Malayalam words. We created the dataset for 26 words where each of these words had been recorded for the above mentioned three types. For each word, 10 EEG samples over 8 channels were recorded. This dataset is useful for developing BCI solutions for patients suffering from neuro-degenerative diseases by developing Machine Learning (ML) classifiers for translating EEG-signals to Malayalam words, vocal or sub-vocal, especially considering the scarcity of datasets available in Indian languages.

A document by laveneswary. Click on the document to view its contents.

The foremost challenge in continual learning is devising strategies to alleviate catastrophic forgetting, thereby preserving a model's memory of prior knowledge while learning new tasks. Knowledge distillation, a form of data regularization, is garnering increasing attention in the field of continual learning for its ability to constrain a model's discriminative power for previous tasks by emulating the outputs of old task models while learning new tasks, thus mitigating forgetting. This paper offers a comprehensive survey of continual learning methods employing knowledge distillation within the realm of image classification. We inductively categorize these methods according to the source of knowledge utilized and provide a detailed analysis of the distillation solutions they employ. Furthermore, given the characteristic inability of continual learning to access historical data, we introduce a novel taxonomy for continual learning approaches from the perspective of auxiliary data usage. In addition, we have conducted extensive experiments on CIFAR-100, TinyImageNet, and ImageNet-100 across nine knowledge distillation-integrated continual learning methods, deeply analyzing the role of knowledge distillation in different continual learning scenarios to alleviate model forgetting. Our substantial experimental evidence demonstrates that knowledge distillation can indeed reduce forgetting across most scenarios.

This paper introduces the QLSTM-DLEM model, a quantum model combining Quantum Long Short Term Memory (QLSTM) and Dynamic Land Ecosystem Model (DLEM), for predicting nitrous oxide (N2O) emissions in agricultural fields. Given the time series data collected from different bifurcation branches, integrating all the data into a single model becomes challenging. To address this, time series data, accompanied by signatures, is utilized to distinguish data from different branches. The auto-correlation function (ACF), partial auto-correlation function (PACF), data distribution, and joint distribution of the first-order difference illustrate that time series data collected under different initial conditions and parameters exhibit slight differences but share several similar characteristics. Performance evaluation metrics, including mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R 2), are employed to assess the model. The results indicate that the QLSTM-DLEM model outperforms the classical LSTM-DLEM model in terms of generalization and stability, possibly attributed to the parallelism and superposition of quantum computing. Overall, our simulation results demonstrate that the proposed QLSTM-DLEM hybrid model can effectively predict N2O emissions in agricultural fields, with potential applications in environmental monitoring and management.

Sustainable resource management within a cloud computing environment is a highly critical and prominently studied research topic. In this context, this paper presented a comprehensive survey of potential sustainable resource management (Sus-RM) strategies that have addressed the energy optimization challenges during cloud workload scheduling and resource management. The perspective of sustainable cloud resource management followed by a discussion of intended motivation, challenges, objectives, and approaches is manifested. The designed research methodology with the proposed method-centric classification and taxonomy of Sus-RM approaches is conferred. Based on the common features of managing sustainability dealing with common operations including task scheduling, virtual machine (VM) placement, and VM rescheduling or migration, the Sus-RM strategies are further grouped into a common class or category. The concept behind each methodbased approach and respective state-of-the-art strategies belonging to each category are concisely discussed with their pandect comparative summary. Besides, conceptual and theoretical analysis, the critical takeaways and lessons learned outlining each method are presented. Further, the trade-off among the leading strategies is capsuled and discussed respectively to their critical features to put forward imperative concluding remarks about the holistic study of cloud Sus-RM. Finally, the scientific survey study is concluded with insightful and concrete future research directions.

Human Abnormal Behavior Detection is needed in several important sectors, especially in the field of public safety. Video Surveillance Systems are a trigger in detecting strange human behavior. However, this abnormal behavior data is rare and requires more costs to obtain, so the unsupervised learning method is used to study normal human behavior patterns only. There have been several previous studies that have provided significant results using this technique. There have been several previous survey studies related to human abnormal behavior detection, but they did not focus on unsupervised learning methods and the scope was too broad. In this paper, I present a survey regarding human abnormal behavior detection in more depth conceptually and classify it into two large parts, such as reconstruction-based detection and generative-based detection. This paper also offers a more extensive comparison with several popular datasets that are often used in previous research. Finally, I also shared several challenges and open research issues that emerged from our survey regarding future directions for future development.

Machine learning models are increasingly vital for clinical decision making. However, the integration of these models into health information systems (HISs) poses significant challenges, mainly due to the disparate formats of electronic health records (EHR) in various healthcare facilities. In response to this challenge, we proposed Model Cabinet Architecture (MoCab), a framework designed to leverage fast healthcare interoperability resources (FHIR) as the standard for data storage and retrieval when deploying machine learning models across various HISs. MoCab simplifies the deployment process by importing and configuring saved prediction models, facilitating patient data retrieval from FHIR servers, and transmitting it to the prediction model for analysis. The framework further incorporates Clinical Decision Support (CDS) Hooks for issuing clinical alerts and uses Substitutable Medical Apps Reusable Technologies (SMART) on FHIR to develop Apps for displaying laboratory results. Finally, MoCab offers the ability to continuously fine-tune and enhance the performance of deployed models over time. We demonstrate MoCab's efficacy through the successful integration of three model types, highlighting its potential in streamlining decision making amidst heterogeneous EHRs. Our proposed MoCab framework not only promotes the reusability of machine learning models across multiple EHRs but also contributes to improving clinical decision making, offering a promising solution to the challenges of integrating machine learning models into healthcare settings.

Bengali is one of the world's most widely spoken languages. Due to its 49 alphabets, this language poses a number of challenges in Optical Character Recognition (OCR). Identifying handwritten graphemes is one of them. Now, the grapheme root, vowel diacritic, and consonant diacritic are the three components of the Bengali grapheme. Therefore, we address this topic by treating it as a multi-label classification problem. In this study, we used the MobileNet architecture. Also, the model we designed performs really well.

With the rise of social networking sites (SNS), a growing concern has emerged regarding the unintentional disclosure of sensitive information, also known as information leakage. This study investigates the causes and consequences of information leakage on SNS through a simulation-based approach. A virtual environment replicating user interactions and data sharing patterns on popular platforms was developed, simulating various scenarios involving unintentional data disclosure, malicious actor activity, and data breaches. The results revealed that oversharing and weak privacy settings were the most common causes of information leakage, while social engineering attacks and data breaches also posed significant risks. The study highlights the importance of user education, improved platform security, and robust privacy controls to mitigate the risks associated with information leakage on SNS.

As various deep learning applications have been deployed in diverse areas, the explainability of neural networks is becoming increasingly important in the research field. Besides being desirable on its own account, explainability also often helps further improve performance of deep learning models. In this work, we introduce float neurons and fixed neurons to describe the neuron-level stability in a network based on the activation pattern of neurons on given input. With the proposed concept, we quantify the expressive ability and robustness of a neural network with a neuron entropy metric and illustrate their relationship by decomposing the computational graph of a neural network. We find theoretically that networks with better generalization have more diverse activation patterns across the input space, which results in a higher neuron entropy globally. On the other hand, the prediction of neural networks is prone to be affected by perturbation when there are locally more float neurons, which respond with additional impulses to local stimuli. Empirically, we show that the proposed analytical framework can be applied to downstream applications, including network pruning and randomized smoothing of network prediction.

The AI-based technologies used in healthcare systems have witnessed significant growth and innovation, as this growth is attributed to innovations in AI and rise in data collection in the healthcare sector. This survey paper provides a comprehensive overview of the diverse technological advancements reshaping the healthcare landscape. The reviewed topics include Medical Image Interpretation using Deep Learning, Generative AI-based Large Language Models (LLMs), Natural Language Processing for Healthcare Records to give a sense of what AI based systems look like in healthcare. For each of these topics, we've delved into their technical aspects and their applications. Through an overview of these cutting-edge technologies, this research aims to shed light on their current state, challenges, and potential implications for the future of healthcare. From enhancing diagnostics to improving patient care and accessibility, AI is poised to play pivotal roles in shaping the healthcare industry for years to come. Furthermore, this survey also delves into the ethical considerations surrounding these technologies.

This paper investigates the Diophantine equation p r + (p + 1) s = z 2 Where p > 3, s ≥ 3 , z is an even integer. The focus of the study is to establish rigorous results concerning the existence of solutions within this specific parameter space. The main result presented in this paper demonstrates the absence of solutions under the stated conditions. The proof employs mathematical techniques to systematically address the case when the prime p exceeds 3, and the exponent s is equal to or greater than 2, while requiring the solution to conform to the constraint of an even z. This work contributes to the understanding of the solvability of the given Diophantine equation and provides valuable insights into the interplay between prime powers and the resulting solutions.

This paper explores the realm of intelligent systems through an analogy inspired by RLC circuits, delving into the interconnected dynamics of reasoning, learning, and control. Leveraging the simplicity and clarity of the analogy, we navigate the conceptual landscape by drawing parallels between electrical components and the cognitive functions of modern AI. The presented analogical framework is the conclusion of the personal experiences of the author in developing intelligent systems, sparked by conversations with fellow researchers and students and presentations of research outcomes. It is worth recognizing the limitations of this analogy, as its reductionist nature may simplify the complexities inherent in intelligent systems. Yet, this exploration provides a fresh perspective on the foundational components of intelligent systems through the lens of the well-established RLC circuit theory.

The implementation of a big data mart using NoSQL technologies, in particular the key-value store, is investigated in this report. It examines the relationship between modern database management systems' benefits and traditional data warehousing methods. Based on actual experience using the Oracle NoSQL Database, the process of going from conceptual schema to logical models is detailed, and its benefits and drawbacks are noted.

A key component of cybersecurity defense is effective information sharing, which promotes teamwork and quick reaction to emerging threats. However, putting strong information-sharing principles into practice might be difficult for practitioners. This summary paper explores cybersecurity specialists’ perceptions on information sharing during real-world international cybersecurity exercises (CDX) and looks at the barriers to the development of better information-sharing skills. The study intends to shed light on the difficulties faced by cybersecurity professionals and pinpoint areas where information-sharing procedures could be improved.