Please see below a demonstration of the latest improvements to EDeN!
This shows in real time a roughly 15x improvements in performance , executing a simple half adder dataset test.
As per the original paper (https://arxiv.org/abs/2103.15552) we do not use back propagation. But instead train for adaptive neuronal spikes from ‘decaying’ energy from stimulus into different neurotransmitters with genetic encoding of morphology.
If you feel like creating an account, and downloading the builds please signup here: https://eden-portal.co.uk/
(The website will undergo variable changes to improve usability soon!)
For a license please contact me at: contact@optishell.co.uk