Host institution: SUITE5 DATA INTELLIGENCE SOLUTIONS LIMITED, Cyprus
Although “black-box” models, where humans are unable to explain how specific inputs affect the results, are satisfactory in a multitude of Artificial Intelligence (AI) applications, their inherent obscurity entails specific drawbacks: a) The lack of information on how big data is processed prevents us from focusing on the input features that really affect the result, b) a vague understanding on how internal decisions are taken makes algorithms less keen to be improved, c) ethically, decisions based on non-explainable systems affecting humans are not acceptable. Therefore, trustworthy eXplainable AI (XAI) algorithms have emerged based on interpretability aiming at: understandability (to understand the inner functionality of the models), comprehensibility (the ability of representing the learned knowledge close to human interpretation) and trustworthiness (the ability of providing reliable predictions). XAI-enabled big data analytics are of at most interest in a Healthcare 4.0 ecosystem, where not only the efficacy of such solutions is important, but also their interpretability, so as to increase the confidence of the involved stakeholder on the algorithm result.
Doctoral Candidate’s (DC) role: The DC will define new structures/methods for tackling the different requirements of analyses that are linked to anonymized and gender-agnostic wearable data and develop efficient data management techniques. Trustworthy and privacy-by-design AI and big data analytics models will be designed, leveraging on approaches such as Federated Learning (FL), able to satisfy the domain needs for secure and privacy-aware AI execution. Given the nature of the domain, special attention will be given on the AI algorithm robustness employing both modular and flexible. Algorithmic structures able to be executed across the whole cloud continuum, and taking into consideration the security and ethics requirements imposed by the data to be processed will be used. In addition, the research will be based on graph-based and hybrid explainable models. Hence, through better interpretability of results and on how internal decisions are taken, the DC will draw key concepts such as understandability, comprehensibility and trust ability for all involved stakeholders, while targeting at 20% increased analytics performance and 80% increased understanding of outputs compared to the State-of-the-Art (SoA).
|A.1. Living allowance (per month)
|A.2. Mobility allowance (per month)
|A.3. Family allowance (per month if applicable)
Doctoral Candidate (DC) is recruited under a Type A employment contract. The living allowance is a gross amount, including compulsory deductions under national law, such as employer and employee social security contributions and direct taxes. Moreover, the DC will be eligible to receive mobility allowance and family allowance (if applicable) that could also be subject to taxation.