Add example w/ semantic decoding w/ neural field

[maedoc]

This is a use case brainstormed during the brain hack in Marseille w/ someone who's name is ftm forgotten.

• acoustic model turns speech into a set of signals
• signals provided as stimulation to part of a neural field
• local & nonlocal connectivity (and perhaps even higher order coupling) do information processing
• a decoder reads out semantic content a la sentence embedding

If instead of neural field it was a transformer, then the models already exist (e.g. whisper). The question is how to do this with from computational neuroscience first principles.

[i-Zaak]

This is very similar in spirit to the D1.9 demonstrator: https://wiki.ebrains.eu/bin/view/Collabs/sga3-d1-9-poc-simulations-for-validity

[maedoc]

Yeah I meant to ask you to post a link.. thanks!

These kinds of examples fit well within the JAX ecosystem since the NN training can be done simultaneously with the VB model training.