ReFRAMing MachinE Learning via Ordinal Tensors and Deep Quality Diversity


FRAME aims to reframe the way machine learning (ML) is theorised and algorithmically realised nowadays. This vision is operationalised through the following three core research objectives.

Obj. 1: From vectors to tensors. FRAME introduces tensor-based ML models able to exploit multi-linear data structural information of extremely high dimensional spaces and yield compact yet highly informative representations.

Obj. 2: From objective to subjective. FRAME expands the current objective ML focus to subjectively-defined tasks that require human input such as emotion recognition. FRAME fuses ordinal tensors with deep preference learning to learn more from less data and train ML models effectively from fewer human demonstrations.

Obj. 3: From quality to deep quality diversity. FRAME offers unconventional solutions to deceptive problems through artificial surprise as a form of search. Deep tensors, human preferences and surprise are uniquely combined to derive diverse outcomes of high quality.

idgThe University of Malta is the highest university in Malta. There are some 10,000 students including over 750 foreign/exchange students from nearly 80 different countries. The Institute of Digital Games was established in 2013, within the University of Malta, and houses a multi- disciplinary research group with backgrounds from Computer Science, the Arts, and the Humanities. The institute is (or has been) involved in several national and international research projects. Institute’s members are involved in the IEEE Computational Intelligence and Games Society, and are members of the editorial boards of the most prestigious journals in the areas of games research and AI. The Institute has established collaborations in the proposed research topic with numerous academic institutes as a result of previous and current research activities in the framework of international and national projects.


There are no publications yet.


There are no news yet.