October 19 ~ 20, 2024, Sydney, Australia
Tadashi Ogino, Department of Information Science, Meisei University, Tokyo, Japan
SHONAN, an advanced system harmonizing human capabilities and information technology (IT), was introduced in light of the COVID-19 pandemic, which prompted the shift of office workers and school students to online platforms. In addition, a specific application of SHONAN, referred to as the narrow area communication system (NAMI), was previously implemented, exclusively sharing text-based information. Further, NAMI uses Bluetooth low energy (BLE) to exchange messages; however, it is difficult to exchange large data, and therefore, we have confirmed that it is possible to exchange large data using Wi-Fi. Thus far, all experimental systems have been designed on paper in advance. This is insufficient for actual dynamic systems. In this paper, we considered a method that can allow NAMI functions to continue to be used even when devices and edges move and the network configuration changes dynamically. Further, we implemented and confirmed the functions in the prototype.
IoT, Sustainable System, Multimedia Data, Autonomous Configuration, Ad Hoc Network.
Maryam Solaiman1, Theodore Mui1, Qi Wang2, Phil Mui3, 1Aspiring Scholars Directed Research Program Fremont, USA, 2University of Texas Austin Austin, Texas, 3Salesforce San Francisco, USA
We model unlearning by simulating a Q-agent (using the reinforcement learning Q-learning algorithm), representing a real-world learner, playing the game of Nim against different adversarial agents to learn the optimal Nim strategy. When the Q-agent plays against sub-optimal agents, its percentage of optimal moves is decreased, analogous to a person forgetting (“unlearning”) what they have learned previously. To mitigate the effect of this “unlearning”, we experimented with modulating the Q-learning so that minimal learning occurs with untrusted opponents. This trust-based modulation is modeled by observing opponent moves that are different from those that a Q-agent has learned. This model parallels human trust which tends to increase with those whom one agrees with. With this modulated learning, we observe that a Q-agent with a baseline optimal strategy is able to robustly retain previously learned strategy. We then ran a three-phase simulation where the Q-agent played against optimal agents in the first phase, sub-optimal agents in the second “unlearning” phase, and optimal or random agents in the third phase. We found that even after unlearning, the Q-agent was quickly able to relearn most of its knowledge about the optimal strategy for Nim.
Reinforcement learning, Q-learning, Nim Game, Unlearning, Learned Memory, Misinformation.
Asmaa EL Harat, Jihad Kilani, Hicham Toumi Youssef Baddi, STIC Lab, FSJ,UCD, EL Jadida, 24000, Morocco.
This survey delves into the complex realm of Internet of Things (IoT) security, highlighting the urgent need for effective cybersecurity measures as IoT devices become increasingly common. It explores a wide array of cyber threats targeting IoT devices and focuses on mitigating these attacks through the combined use of deep learning and machine learning algorithms, as well as edge and cloud computing paradigms. The survey starts with an overview of the IoT landscape and the various types of attacks that IoT devices face. It then reviews key machine learning and deep learning algorithms employed in IoT cybersecurity, providing a detailed comparison to assist in selecting the most suitable algorithms. Finally, the survey provides valuable insights for cybersecurity professionals and researchers aiming to enhance security in the intricate world of IoT.
Internet of Things (IoT), cybersecurity, machine learning, deep learning.
Maryam Solaiman1 and GM Solaiman2. 1Aspiring Scholars Directed Research Program Fremont, USA, 2Cisco Systems, Inc., San Jose, USA
Since their conception in the early 1970s, microprocessors have been put to a multitude of uses through various dif erent designs. While there are many academic papers on the implementation of a microprocessor, only a few are devoted to verification. Design and verification go hand in hand with every stage of a digital circuit implementation. In this paper, we propose a RISC pipelined processor with hazard detection, automatic hazard resolution, and automatic stall insertion. We incorporate a modular approach to design. We proposed a constrained random verification environment to fully verify the design with coverage based verification. Finally we implemented the processor in real hardware to demonstrate operational ability. Our approach could easily be scaled up to design, verification and implementation of a large scale system on chip manufacturing.
MIPS32, microprocessor, hazard detection, Verilog HDL, verification, coverage, computer architecture.
Sonjoy Ranjon Das, Department of Computer Engineering, Northumbria University, London, UK
Lung Cancer (Lc) Presents a Critical Global Health Challenge, Requiring Rapid and Accurate Diagnosis for Effective Treatment. Traditional Diagnostic Methods Often Fall Short in Precision, Leading to Delays. This Study Evaluates the Performance of Six Transfer Learning Models—mobilenetv3, Densenet201, Efficientnetb7, Vgg16, Vgg19, and Inception V3—in Predicting Lung Cancer Using a Dataset of 15,000 Histopathology Images. The Models Classify Lung Cancer Types, Including Adenocarcinoma, Benign, and Squamous Cell Carcinoma. Mobilenetv3 Emerges as the Most Efficient, Achieving 99.70% Accuracy, Outperforming Inception V3 (78%), Densenet201 (93%), Vgg16 (99%), Vgg19 (98%), and Efficientnetb7 (99.50%). Evaluation Metrics Such as Accuracy, Precision, Recall, and F1-score Indicate That Mobilenetv3 and Efficientnetb7 Offer Superior Performance. The Study Suggests These Two Models as the Best Options for Lung Cancer Classification.
Deep Learning, Lung Cancer Prediction, MobileNetV3, VGG16, VGG19, InceptionV3, DenseNet201, EfficientNetB7, SoftMax layer, CT images.
Aeshna Kapoor, Lead Data Scientist, BNY Mellon, New York, USA
The rapid evolution of data-driven technologies has led to the proliferation of big data systems capable of managing and analyzing vast amounts of data. However, many organizations continue to rely on legacy systems that are deeply entrenched in their operations. The challenge lies in integrating these legacy systems with new, AI-driven platforms to create a cohesive, hybrid infrastructure that leverages the strengths of both. This paper presents a comprehensive approach to designing and implementing a hybrid big data infrastructure that combines legacy systems with advanced AI technologies. We explore the challenges, architectural considerations, and the potential benefits of such an integration, aiming to provide a roadmap for organizations seeking to modernize their data infrastructure without completely abandoning their existing investments.
Big Data, Hybrid Infrastructure, Legacy Systems, AI Integration, Data Platforms.
Swapna Krishnakumar Radha, Andrey Kuehlkamp, and Jarek Nabrzyski, Center for Research Computing, University of Notre Dame,Notre Dame Indiana, USA 46556
Attestation of documents like legal papers, professional qualifications, medical records, and commercial documents is crucial in global transactions, ensuring their authenticity, integrity, and trustworthiness. Companies expanding operations internationally need to submit attested financial statements and incorporation documents to foreign governments or business partners to prove their businesses and operations’ authenticity, legal validity, and regulatory compliance. Attestation also plays a critical role in education, overseas employment, and authentication of legal documents such as testaments and medical records. The traditional attestation process is plagued by several challenges, including time-consuming procedures, the circulation of counterfeit documents, and concerns over data privacy in the attested records. The COVID-19 pandemic brought into light another challenge: ensuring physical presence for attestation, which caused a significant delay in the attestation process. Traditional methods also lack real-time tracking capabilities for attesting entities and requesters. This paper aims to propose a new strategy using decentralized technologies such as blockchain and self-sovereign identity to overcome the identified hurdles and provide an efficient, secure, and user-friendly attestation ecosystem.
Attestation, Blockchain technology, Self-sovereign Identity technology.
Olalekan M. Durojaiye, Ramanjit K. Sahi, Department of Mathematics & Statistics, Austin Peay State University, TN, USA
The loan data simulated with Monte Carlo approach and analyzed in the research work provides valuable insights into the borrowers’ financial positions and loan performance. By calculating the debt-to-income ratio (DTI), we identified 122 (50.8%) loans that were at high risk of default. We also used risk-based pricing (RBP) to assign higher interest rates to riskier loans, helping to mitigate the risk of default. The data analysis showed that a higher DTI is associated with a higher risk of default, and a higher RBP is associated with a higher interest rate. Therefore, it is essential to use these metrics when assessing loan applications to ensure a healthy loan portfolio. This analysis can be used to inform loan officers, risk analysts, and other stakeholders involved in the lending process.
Loan Simulation, Interest rate, Risk Mitigation, Debt-To-Income Ratio, Risk-Based Pricing.
Houssam Hamici , Hani Ahmad and Hamido Hourani, Department of Electrical Engineering, PSUT, Amman, Jordan
This work presents a brain tumor classification survey utilizing Convolutional Neural Networks and Vision Transformers methods. The classification is based on a brain tumor dataset comprising 7023 MRI images with four classes present in the dataset: No tumor, Pituitary, Glioma, and Meningioma. The models used for the classification are ResNet101V2, VGG19, MobileNetV2, InceptionV3, Xception, and Vitb16. Two types of experiments were conducted with and without pre-trained weights to classify the dataset with intended models. Since the dataset is relatively small, CNN performs better than ViT since ViT relies on a vast pre-trained dataset to perform very well. The best results were obtained by Inceptionv3 and Xception architectures, both achieving an accuracy of around 98.6%.
BrainTumor(BT), Classification, Convolutional Neural Networks (CNN), Vision Transformers (ViT).
Arijit Das, Tanmoy Nandi, Diganta Saha, Department of Computer Science and Engineering, Jadavpur University,Kolkata, West Bengal, India
This paper presents a novel approach to predicting financial market trends by inte-grating deep learning models with natural language processing (NLP) techniques applied to Twitter data from influential leaders. Unlike traditional models reliant solely on historical fi-nancial data, our method leverages real-time social media information to enhance predictive accuracy. Key contributions include the development of a versatile algorithm capable of gener-ating models for any Twitter handle and financial component, as well as predicting the tem-poral window during which tweets affect stock prices. We also explore the combined effects of multiple Twitter handles on trend prediction. Through a comprehensive survey, we identify research gaps, collect necessary data, and propose a state-of-the-art algorithm with a complete implementation environment. Our results demonstrate significant improvements facilitated by NLP analysis of Twitter data on financial market components. We focus on the Indian and USA financial markets, with potential for extension to other markets. In conclusion, we discuss the socio-economic implications and utility of our approach in informing decision-making processes within financial markets.
Deep Learning, Financial Market Prediction, Twitter Analysis.
Jesús María Velásquez-Bermúdez, Founder and Chief Scientific Officer, HYPOTHALAMUS Artificial Intelligence Inc., USA
The document explores the advanced integration of Artificial Intelligence within Enterprise Optimization Systems. The core innovation presented is the transition from traditional Decision Support Systems (DSS) to Enterprise-Wide Optimization Systems (EWOS), which are designed to optimize organizational decision-making processes holistically and autonomously. The Enterprise Artificial Brain concept, inspired by the human brains structure, incorporates artificial components like the neocortex, hypothalamus, and hippocampus to manage, produce, and store knowledge. This integration allows for Autonomous Real-Time Distributed Optimization, significantly enhancing the efficiency and effectiveness of business operations. The document further discusses the application of these principles in various industrial contexts, particularly in the oil and gas sector. HAI’s research underscores the evolution from mental planning models to sophisticated, mathematical optimization models, facilitating integrated business planning/scheduling, and decision-making. By employing technologies such as OPTEX, Optimization Expert System, a generative AI system, HAI demonstrates how artificial brains can autonomously manage complex industrial processes, thereby reducing development time and increasing decision-making accuracy. This approach aims to emulate human cognitive functions through artificial mathematical systems, providing organizations with robust tools for navigating dynamic and uncertain environments.
Bringing these components together enables HAI to start thinking about a revolutionary idea: the Artificial Brain.
Ravi Kumar and Ayushi Kumari, Department of Computer Science and Engineering, Arya College of Engineering and Research Center, Jaipur, Rajasthan India
The integration of Natural Language Processing (NLP) in chatbot technology has revolutionized the healthcare sector, offering innovative solutions for patient care and management. This paper explores the diverse applications of advanced NLP-based chatbots in smart healthcare. These applications include providing medical information, assisting in disease diagnosis, supporting mental health, managing chronic diseases, and enhancing patient engagement. We discuss the underlying technologies, benefits, challenges, and future directions for NLP-based healthcare chatbots.
Natural Language Processing, Smart Healthcare, BERT, Chatbots, Mental Health, Sentiment Analysis.
Zeyu Zhang1, Yu Sun2, 1Santa Margarita Catholic High School, 22062 Antonio Pkwy, Rancho Santa Margarita, CA92688, 2Computer Science Department, California State Polytechnic University, Pomona, CA91768
This research paper presents the development and evaluation of a personalized mental health support applicationthat leverages AI-driven features for real-time user interaction [1]. The application includes components for facial expression classification and real-time image generation, both of which were subjected to rigorous testing throughtargeted experiments [2]. The first experiment evaluated the accuracy of the emotion recognition system, revealingstrong performance with distinct emotions but highlighting challenges with subtle expressions. The secondexperiment tested the responsiveness of the image generation component, showing ef ective performance with simpleinputs but identifying delays with more complex tasks. While the application demonstrates significant potential, especially in its ability to provide tailored emotional feedback and support, further refinement is needed to enhanceaccuracy, performance, and data security. The findings suggest that with continued development, this applicationcould become a valuable tool in the field of mental health and emotional well-being.
Facial Expression Classification, AI-Driven Mental Health, Real-Time Image Generation, Emotion Recognition, Emotional Well-Being.