Research
Dendroprop: Learning to link somatic and dendritic inputs over long timescales
A two-compartment neuron model in which NMDA-like plateau potentials link dendritic and somatic spike patterns, trained with a new surrogate-gradient rule.
What we research
We study how neurons process information to build the next generation of energy-efficient and powerful AI computers inspired by the brain's example.
→ More about usGet in touch