Compensatory Hebbian learning for categorisation in simulated biological neural nets

We've been working on AI systems based on simulated neurons for about 15 years now.

We've done some prior work using compensatory Hebbian learning for categorisation, but we couldn't move beyond the sensory interface with learning.

We've recently extended our neural model, so that hypo-fatigue can lead to spontaneous neural firing.

Using a novel form of compensatory learning, this enables the deep structure to categorise.

The paper, and thus this talk, is based on the task of categorising yeast.