Learning From Data Streams
This project focuses on the development of novel machine learning approaches that can learn from stream data collected from non-stationary environments and make decisions under harshest learning conditions. We tackle various challenges including noisy data, high-dimensional data, and missing data, learning from skewed-class distributions with the rarity of labeled data, novelty detection, and handling extreme verification latency.
Explainable Artificial Intelligence
Explainable Artificial Intelligence (XAI) tries to produce explainable models enabling users to appropriately trust and interpret the attained results and reveal the model functionalities. There has been an increasing need for XAI in various applications, such as cyber-physical energy systems, autonomous vehicles, and healthcare. This project aims to develop explainable artificial intelligence models and assessment criteria for smart energy management, prognostic and health management in cyber-physical energy and power systems.
Attack-Resilient and Fault-Tolerant CPS
The welfare and security of modern societies rely on the safe and secure operation of complex safety-critical cyber-physical systems (CPSs). CPSs dependency on digitalization, wireless communication, and remote control systems increases their vulnerabilities to malicious threats, which lead to the loss of system integrity and functionality. This project focuses on integrating the knowledge on machine learning, big data analytics, cybersecurity, and cybernetics that would pave the way together towards fault-tolerant attack-resilient CPSs.
Secure Cybernetics in Industrial CPS
Nowadays, almost every aspect of technology—from mobile devices to smart grids and multi-agent control systems—is impacted by the integration of computational intelligence with the communication systems networks. This phenomenon has brought about a vast variety of challenges to modern cyber-physical systems from a security viewpoint. The objective of the proposed research is to discover novel methods to secure computational models during construction, calibration, and communication. The outcome of this research will guarantee the safe and secure operation of the new generation of dynamic intelligent systems that are more vulnerable to malicious cyber-activities.
Missing Data Analysis
Missing data are inevitable in almost all industries and highly undesirable in machine learning, data mining, and information systems. There exist a number of reasons for this severe deficiency, including imperfect procedures of manual data entry, incorrect measurements, data collection problems, and equipment errors.
This project focuses on the development of efficient algorithms for the treatment of missing data in order to improve decision-making.