Aditya Gilra

The paper shows that prediction-error signals in the anterior cingulate cortex are necessary for rapid task-switching in mice, with specific neural circuits involved in computing these signals.

BunDLe-Net learns low-dimensional Markovian embeddings of neural dynamics that preserve only behaviourally relevant aspects, improving the relationship between neural activity and behaviour across datasets.

The paper develops a thalamocortical network model to explain novelty detection through synaptic depression in cortical layer 6 and thalamic oscillations, showing similarity to EEG-based mismatch negativity.