PhD Student

Last application date
Oct 08, 2021 00:00
TW05 - Department of Information Technology
Limited duration
computer science, biomedical engineering, or similar
Occupancy rate
Vacancy type
Research staff

Job description

A PhD position with grant is available at the GASP interdisciplinary research group at UGent. The PhD is connected to a project, "SoundStreams", that has as an objective to further develop our novel cognitive architecture for artificial intelligence (AI) with improved capabilities for continual, one-off learning in dynamic contexts with reduced risk of catastrophic forgetting, and to improve the representation (explicit or implicit) of time, admitting predictive coding models as a basis for learning. It will do so by translating recent findings on human auditory perception and learning to interpretable AI sub-models at an algorithmic level. Three main topics will be addressed: (a) Concept representation focusing on hierarchy in scope and time will be explicitly implemented. It will foster concurrent grouping for creating events (e.g. a car), sequential grouping for connecting them to expected streams (e.g. traffic), and segmental grouping for grouping streams into perceptual units (e.g. morning rush hour). (b) Attention (selective processing/gating of stimuli based on information content, prediction mismatch, and internal or external orientation of cognitive resources) will be developed through implicit modelling of focusing and gating depending on the instantaneously most appropriate system state. (c) Learning states as a selective mechanism for different types of learning and memory consolidation will be inspired by the coordinating role for the central nervous system played by the Locus Coeruleus - Norepinephrine axis in humans. Four states will be introduced: sleep, quiet wake, active wake, and fight-or-flight, each corresponding to a particular level of sensory processing and learning. Proof-of-concept of the superior behavior will be given for learning from a continuous auditory stream that includes significant changes in context.

Job profile

  • Candidates should preferably have a degree in computer science (artificial intelligence) with strong interest in human auditory perception and cognition or a degree in biomedical engineering with a strong interest in artificial intelligence.
  • Candidates with other university degree and prior experience in this research field are also encouraged to apply.

How to apply

Apply by email to Applications should be accompanied by a full cv, motivation letter, and transcript of master and bachelor degree clearly mentioning the degree obtained and the ranking amongst peers.