Trustworthy and Explainable Machine Intelligence – RELATE

Research Activity

RELATE (tRustworthy ExplainabLe mAchine inTElligence) a multidisciplinary group of researchers and technologists, who conduct fundamental and applied research through synergistic cooperation and borderless exchanges within and across several disciplines, including data mining, machine/deep learning, computational intelligence and high-performance computing. Specifically, RELATE focuses on intelligent knowledge-based systems that can learn from volumes of data and accordingly reason for pattern discovery and analysis, predictions, and decision making in scientific as well as applicative domains.

The main activities are related to:

• Machine intelligence models and algorithms. This activity aims to design interpretable approaches and methodologies for optimizing and automating processes, extracting and classifying data, detecting, analyzing and predicting trends/patterns and enhancing interaction humans/environment, through safer, fairer and more readable models and algorithms, such as for example, decision trees, decision rules and linear regression;
• Security intelligence techniques. The objective is to conceive and develop intelligent methodologies and techniques to deal with security and privacy issues through trustworthy and interpretable approaches. The use of technologies as cyber-physical security (Physical Unclonable Functions) and Blockchain/DLT, is also an important side of this activity;
• Modeling & simulation of complex systems. The purpose is to design and implement interpretable models and algorithms for natural disaster prevention, also exploiting high-performance scientific computing;
• Green and sustainable development in Smart Environments, also assuring that the methodologies and the algorithms adopted are behaving responsibly and trustworthy, is the aim of this activity.

Goals

RELATE aims to develop innovative intelligent solutions featured by a high degree of trustworthiness and intelligibility. In particular, trustworthiness is pursued via advanced developments overcoming the known limitations of intelligent systems (such as, e.g., the vulnerability to attacks, the lack of privacy protection, and the bias against under-representations), while Intelligibility is pursued to provide humans with clear, detailed, and insightful explanations of operations and results of machine intelligence techniques. Besides, as many of the targeted applications can be computationally demanding and critical, the designing of intelligent systems for such scenarios is expected to exhibit scalability properties, adaptation to ever-changing environments, and fault tolerance.

Application Fields

Cybersecurity, Healthcare, Industry 4.0, HPC