Humera Batool and Zhonggui Sun, School of Mathematical Sciences, Liaocheng University, Liaocheng 252000, China
Pandemic 2019-nCoV effects humans and animals. Human life, health and economy is effected, including almost all countries in World. As its a viral disease, various countries have prepared vaccines for it which vary in effectiveness. Lungs are worst effected in this disease. Lungs related problems are diagnosed with imaging techniques like chest CT and X-ray(radiography). As compared to chest CT scan, chest X-ray is quite cost effective. Deep learning is sublime technique of machine learning that can aid in screening COVID-19. To examine COVID-19 on X-ray images by machine learning methods, we proposed the approach to classify chest x-ray images using cross vision transformer (Cross-ViT), that proved better results over existing state-of-the-art image classification models including CNN, Inception V3, ResNext, ResNet, Xception, Vision Transformer. Out of all these models, Cross ViT gave highest accuracy.
Saif Hussein Abdallah Alghazo, Norshima Humaidi* and Shereen Noranee, Registration Authority, Abu Dhabi Global Market, Al Maria Island, Abu Dhabi, United Arab Emirates, Faculty of Business and Management, Universiti Teknologi MARA (UiTM) Selangor, Malaysia
Cybersecurity threat become a serious issue recently, which is usually constituted by people carelessness, ignorance and failure to practice cybersecurity behaviour adequately. Using a data from a quantitative survey, Partial Least Squares-Structural Equation Modelling (PLSSEM) analysis was used to determine the factors that affect cybersecurity protective behaviour (CPB). This study adapts cybersecurity protective behaviour model by focusing on two constructs that can enhance CPB: information security competencies (ISI) of the security leaders and procedural information security countermeasure (PCM) awareness. Theory of leadership competencies were adapted to measure user°«s perception towards competencies among security leaders in the organization. Confirmatory factor analysis (CFA) testing shows that all the measurement items of each constructs were adequate in their validity individually based on their factor loading value. Moreover, each constructs are valid based on their parameter estimates and statistical significance. The quantitative research findings show that PCM awareness strongly influences CPB compared to ISI. Meanwhile, ISI was significantly PCM awarenss. This study believes that the research findings can contribute to human behaviour in IS studies and are particularly beneficial to policy makers in improving organizations°« strategic plans in information security, especially in this new era. Most organizations spend time and resources to provide and establish strategic plans of information security; however, if employees are not willing to comply and practice information security behaviour appropriately, then these efforts are in vain.
Cybersecurity, Protection Behaviour, Information Security, Information Security Competencies, Information Security Awareness.
Luiza Naschon, Senior Security Engineer, Red Hat, Israel, Martin Ukrop, Senior Technical Program Manager, Red Hat, Brno
OpenShift uses Route objects to expose web applications to the outside world through HAproxy. One of the challenges of managing web application routing in containerized environments such as OpenShift is how to securely transfer information and allow access to the applications running in those environments. In this paper, we will go through two possible attacks discovered during the security research on OpenShift networking, the route takeover, and DNS hijacking. Then we will present and explain how the users can prevent those possible attacks by following security practices.
Networking, Routes, Containerized Network, Hijacking, Network Security Policies, Route Takeover.
Bob Zigon1 and Fengguang Song2, 1Beckman Coulter, Indianapolis, IN, USA, 2Department of Computer Science, Indiana University-Purdue University Indianapolis, Indianapolis, IN, USA
Mnihs seminal deep reinforcement learning paper that applied a Deep Q-network to Atari video games demonstrated the importance of a replay buffer and a target network. Though the pair were required for convergence, the use of the replay buffer came at a signiffcant computational cost. With each new sample generated by the system, the targets in the mini batch buffer were continually recomputed. We propose an alternative that eliminates the target recomputation called TAO-DQN (Target Accelerated Optimization-DQN). Our approach focuses on a new replay buffer algorithm that lowers the computational burden.We implemented this new approach on three experiments involving environments from the OpenAI gym. This resulted in convergence to better policies in fewer episodes and less time. Furthermore, we offer a mathematical justification for our improved convergence rate.
DQN, Experience Replay, Replay Buffer, Target Network
Junfei Zhang1, Yueqi Li2, and Xinbo Zhou3, 1School of Computing and Information Systems, The University of Melbourne, Melbourne, Victoria, Australia, 2College of Physical Science and Technology, Xiamen University, Xiamen, Fujian, China, 3Faculty of Information Technology, Beijing University of Technology, Beijing, China
Prediction of the electronic structure of functional materials is essential for the engineering of new devices. Conventionalelectronic structure prediction methods based on density functional theory (DFT) suffer from not only high computational cost, but also limited accuracy arising from the approximations of the exchange-correlation functional. Surrogate methods based on machine learninghave garnered much attention as a viable alternative to bypass these limitations, especially in the prediction of solid-state band gaps, which motivated this research study.Herein, we construct a random forest regression model for band gaps of double perovskite materials, using a dataset of 1306 band gapscomputed with theGLLBSC (Gritsenko, van Leeuwen, van Lenthe, and Baerendssolid correlation)functional. Among the 20 physical features employed, we find that the bulk modulus, superconductivity temperature, and cation electronegativity exhibit the highest importance scores, consistent withthe physics of the underlying electronic structure. Using the top 10 features, a model accuracy of 85.6% with a root mean square error of 0.64 eV is obtained, comparable to previous studies. Our results are significant in the sense that they attest to the potential of machine learning regressions for the rapid screening of promising candidate functional materials.
Machine Learning, Random Forest Regression, Electronic Structure, Computational Material Science.
HuiBing Xie1 and Yu Sun2, 1Northwood High School, 4515 Portola Pkwy, Irvine, CA 92620, 2California State Polytechnic University, Pomona, CA, 91768, Irvine, CA 92620
Since some years ago, the popularity of drawing has been increasing. There are a lot of existing tools to help people to improve their drawing . Some tools provide human body images, so people can practice their human body drawing . However, users cannot find the desirable pose images since these tools provide only a list of images but it cannot be sorted by pose. Thus, we proposed a tool in which users can move the joints of a stick figure to obtain the matching human pose image. In our experiments, the result shows that the engine matches 87% of the human images and the stick figure. Also, we performed data analysis with feedback from 10 high school students. The result shows that 5 out of 10 students were satisfied with our tool.
MediaPipe, Pose Estimate, Drawing, Matching.
Simisani Ndaba, Edwin Thuma and Gontlafetse Mosweunyane, Department of Computer Science, University of Botswana, Gaborone, Botswana
In this paper, an investigation is done to identify writing style features that can be used for cross-topic and cross-genre documents in Authorship Identification. In particular, we empirically evaluate different writing style features that were previously used in single topic and single genre documents in Authorship Identification to determine whether they can be used effectively for cross-topic and cross-genre Authorship Identification using an ablation process. The dataset used was taken from the 2015 PAN CLEF English collection consisting of 100 sets from the PAN CLEF website. Furthermore, we investigate whether combining some of these feature sets can help improve the authorship identification task. Three different classifiers were used which include Na√Įve Bayes, Support Vector Machine and Random Forest. The results suggests that a combination of lexical, syntactical, structural and content feature set can be used effectively for cross-topic and cross genre authorship identification as it achieved a result of 0.837.
DQN, Experience Replay, Replay Buffer, Target Network.
Andy Jiang1 and Yu Sun2, 1Klein Oak High School, 22603 Northcrest Dr, Spring, TX 77389, 2California State Polytechnic University, Pomona, CA, 91768, Irvine, CA 92620
Since the beginning of the COVID-19 pandemic, education largely shifted away from the physical classroom and towards more digitally oriented platforms. This simplified classroom attendance problems greatly, as newly created programming scripts could easily track the students in a meeting room via their names. However, with the recent growing return to in person education, it has become apparent that the problem of attendance within the context of a non-virtual classroom environment has yet to be solved in an efficacious automated fashion. In larger classrooms, the severity of this problem becomes exacerbated even further, as teachers are forced to allocate valuable time for the purpose of marking attendance. The flourishing world of machine-learning based algorithms were the first solutions that we considered, and within the context of the premise, we concluded that facial recognition would likely be the most feasible and effective approach that we could use. This paper develops a mobile application to apply real time face recognition for the purpose of the above stated problem, using a combined backend of the Firestore database and Amazon AWS services. Applying our application to in person classrooms, the results show that our solutions are immensely effective in both saving time and reducing error.
Machine learning, Flutter, Facial Recognition.
Atef Hadi Ataya, Department of Engineering Management, University of Wollongong, Dubai
Diabetes is understood to be an ailment where the human being systemís blood sugar amounts tend to be unusually higher. Itís acknowledged all over the globe among the long-term problems. Diabetes prevents your bodyís capability to help to make insulin, leading to extreme blood sugar levels as well as gluconeogenesis abnormalities. A lot of women are influenced by gestational diabetes, the industry type of diabetes occurring throughout being pregnant. Ladies tend to be more likely compared to guys to build up diabetes-related difficulties, as well as women that are pregnant may create gestational diabetes throughout their pregnancy. This research chose the well-known Logistic Regression (LgR), k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), Random Forests (RF), XGBoost, and LightGBM, for diabetes prediction. A comparative study of the algorithmic performances is performed to identify the best valuable algorithm in the clinical decisions system.
Artificial Intelligence, Machine Learning, Diabetes, Disease Detection, Healthcare.
Tae Hyun Kim, Won Seok Jan, Sun Cheol Heo, Min Dong Sung and Yu Rang Park, Department of Biomedical Systems Informatics, Yonsei University College of Medicine, South Korea.
Federated learning (FL) has been used for model building across distributed clients. However, FL cannot leverage vertically partitioned features to increase the model complexity. In this study, we proposed a personalized progressive federated learning (PPFL) model, which is a multi-model PFL approach that allows the leveraging of vertically partitioned client-specific features. The performance of PPFL was evaluated using two datasets: the Physionet Challenges 2012 dataset and a real-world dataset composed of eICU data and the Severance Hospital, Seoul, South Korea. We compared the performance of inhospital mortality and length of stay prediction between our model and the FedAvg, FedProx, and local models. The PPFL showed an accuracy of 0.849 and AUROC of 0.790 in in hospital mor-tality prediction, which are the highest scores compared to client-specific algorithm. For length-of-stay prediction, PPFL also showed an AUROC of 0.808 in average which was the highest among all comparators.
Personalized Federated Learning, Vertical Federated Learning, Non-IID data.