Old Categorisation and Vowels
- I've been working on this stuff for a long time.
- We work with Hebbian learning which (roughly) increases the
synaptic weight when the pre and post-synaptic neuron co-fire.
- The first "real" task I did was the congressional voting task
from the UCI repository.
- We did a two-fold test. Trained by presenting a congressman's vote
on 15 bill and his party.
- Testing on the party got about 89%
- The data had missing items, and you could predict the votes knowing
the other features.
- This paper had some stuff I did with Vivianne Orengo on IR with CAs.
- I also did vowell learning more recently; the result is ok, but I'm
clearly not ready for speech recognition yet.
- We also used a compensatory learning rule to learn hierarchies.