Electrical, Informatics Engineering
[Closed]
Work description
Current systems of object representations still fully rely on the powerful capacity of recognizing patterns within the vast space of pixel information; however, addressing object individuation as solely a pattern recognition task has left core notions that are leveraged by humans from early childhood: spatiotemporal rules that seem to rule how objects interact with the environment and between each other. The objective of this project is to develop neurosymbolic systems for learning object representations that are constrained by innate knowledge about the intuitive physical behavior that characterizes rigid objects, implemented by universal probabilistic programs.
Academic Qualifications
The awarding of the grant is dependent on the applicants' enrolment in study cycle or non-award courses of Higher Education Institutions.
Minimum profile required
Experience in deep learning and probabilistic inference.
Preference factors
Experience in research projects and writing of scientific papers.
Application Period
Since 07 Dec 2023 to 21 Dec 2023
[Closed]
Centre
Telecommunications and Multimedia