Cyber Social FML - Computing III. Architectures

• Main Authors: Hahanov, Vladimir, Litvinova, Eugenia, Hacimahmud, Abdullayev Vugar, Chumachenko, Svetlana, Khakhanova, Hanna, Hahanov, Ivan
• Format: Conference Proceeding
• Language: English
• Subjects: Computational modeling; Computer architecture; cyber-social computing; cyber-social ML-computing; Data models; Ethics; Federated Machine Learning; general ML - equation; human-free ML - management; Machine learning; Mathematical models; metrical moral relations; quantum data structures; traditions - legislation difference; Training
• ISSN: 2472-761X
• DOI: 10.1109/EWDTS52692.2021.9581033

A new cyber-social computing of moral relations between the state and citizens is proposed, which play the role of tolerant mechanisms of management and execution to create moral and material values, improve the quality of life and preserve the environment. Computer engineering technologies are being formed, aimed at the harmonious sustainable development of society, excluding social collisions, revolutions and wars, based on metric exhaustive monitoring and moral digital management. An improved model-architecture of cyber-social FML-computing is proposed, which differs from the standard one by combining the process of training, testing, functioning and distribution of ML-terminals in space, which makes it possible to reduce the training time by an order of magnitude and improve the quality of services for recognition and decision-making when serving citizens. An advanced cyber-social computing cloud-edge architecture for federated learning algorithms is proposed, which includes four phases: local training (Training), uploading parameters (Upload) to the cloud model, aggregating parameters on the cloud, and returning model parameters to terminals (Download). A logical metric of the quality of pattern/defects/collision recognition is introduced, which, together with the computing equation, makes it possible to form all structural and normalized assessments in the learning process. A logical metric of the quality of recognition of patterns, defects and collisions is introduced, which, together with the computing equation, makes it possible to form all structural and normalized assessments in the learning process.