opencog-dev team mailing list archive

["Joel Pitt" <joel.pitt@xxxxxxxxx>]

Hi Ben, Sorry for the lack of updates last week, I've been a bit flat
out and juggling between contracts so haven't had a full day of
opencog stuff for a while. Still been doing the work though, it's just
been a little more fragmented than usual...

Anyhow, some things I've been doing:

-Running hopfield network replications - because the link adding is
stochastic any one run may have poor performance. I've created python
scripts to graph the performance after each imprint so that it can be
judged easily (I got tired of looking at text tables!) and also
created scripts to graph the final distribution of performance across
replicates. I'm going to through a few of these up on the wiki tonight
after visiting my Mum for Mother's day (even if it is a slightly
consumerist holiday, it's still nice to spend time with family!).
Smaller (3-5 square) networks have pretty good performance (consistant
improvement for a mutated cue pattern) but 10x10 networks are still a
bit erratic at times... still have ideas, but I'd like to commit what
I've already done to the wiki before I start fiddling with more
aspects attention allocation.

-Wrote a blog article on the ImportanceUpdatingAgent for OpenCog Brainwave.

-Added a bunch of miscellaneous stuff to the OpenCog wiki and
suggested wiki improvements to David about formating, word
definitions, and math (Dave: I saw you added math and source code
support, nice!).  Some things that might be of note when you are
writing/converting the documentation: There is a theory page linked
off of the sidebar, so you'll probably want to add to or link from it
http://www.opencog.org/wiki/Theory . As I mentioned in another email,
there are pages for LinkTypes and NodeTypes with a page for each type.
There is also MindAgents and each MindAgent should have a separate
page too. We've also put in place a tool tips system that allows a
wiki page to act as the text of a tooltip (these pages should begin
with "tt:") - this is useful for difficult words that have reasonably
simple definitions (in particular I'm thinking about the lingo for NLP
- where I hadn't heard things like meronym and holonym before).

-Reading more about PLN - are there any particular chapters in the
book that are more important than others, particular for our meeting
next week? I'm about half way through so far.

Cheers,
J