The rhythm of predictive coding

This project investigates the role of brain oscillations in the mechanisms involved in predictive coding.

The details

Predictive coding is an influential framework of cortical organisation. However, the canonical predictive coding model treats cortical processing as a stationary process: input remains constant and the sensory hierarchy converges on a minimum-error computational solution. Of course, the real world is dynamic and ever-changing, and existing predictive coding models could not handle time-variant input. However, earlier this year, PI Hogendoorn proposed an extension to the canonical predictive coding framework that not only allows predictive coding to process time-variant input but would allow also the network to compensate for the delays that inevitably accumulate during neural transmission (Hogendoorn & Burkitt, eNeuro 2019). This project investigates how this might be achieved at the neural level.

The project is structured as follows:

Year 1: The project commences at UoM, where the graduate researcher will be trained by PI Hogendoorn in the theoretical basis of predictive coding and related concepts and start collecting psychophysical and EEG data.

Year 2: The graduate researcher receives additional training in time-frequency analyses of both behaviour and EEG signals in the lab of PI Landau, and learns to identify the signatures of rhythmic neural mechanisms in those signals as well as carry out further analyses.

Year 3: The graduate researcher will return to UoM to finalise the thesis and integrate the empirical work with the theoretical framework of predictive coding under the guidance of PI Hogendoorn.

Supervision team

University of Melbourne supervisor:
Dr Hinze Hogendoorn

Hebrew University of Jerusalem supervisor:
Associate Professor Ayelet Landau