Generative AI for computational modelling of binocular visual processing
Generative models of binocular visual processing whose outputs are validated, uncertainty-aware and interpretable.
Background
Understanding how the visual cortex integrates information from both eyes is central to vision research, and neuroimaging and EEG signals make it possible to study this process computationally.
Research questions
- How can generative models capture visual cortex activity from neuroimaging and EEG signals?
- How can their outputs be validated and made interpretable?
Objectives
- Develop and evaluate multimodal generative models of visual cortex activity
- Quantify the uncertainty of the model outputs
- Assess the interpretability of the learned representations
Approach
Generative and multimodal modelling with uncertainty estimation and explainability analysis, applied to neuroimaging and EEG signals.
Related experience
- Earlier work on EEG-based epilepsy prediction
- Calibrated general anaesthesia triage model (manuscript under review)
- Time-series and bio-signal analysis
The full proposal is shared privately on request. Request the full proposal · All proposals