Hey Marek, Fergal, Thanks for the suggestions, I will definitely give those a go! Im currently writing my thesis about HTM, which is why there’s a bit of a sense of urgency in my e-mails :). I’m definitely looking forward to getting everything running and starting to experiment with the code.
From a theory perspective, there is a lot of discussions I’d like to open up to the subscribers in due time. Comparisons to ANN, a potential graphical interpretation of the inference stage, forgoing strict hierarchy for a modular design with multiple HTMs connected to an analogue of the thalamus (or maybe an auto associative memory), etc. I’m sure you fellas will have great insight and I’m excited to get your feedback. In the meantime, thanks for all the assistance :). Nicholas On Nov 16, 2013, at 6:29 PM, Marek Otahal <[email protected]> wrote: > Hi Nicolas, > > sorry for a late reply. > > > On Sat, Nov 16, 2013 at 2:42 PM, Nicholas Mitri <[email protected]> wrote: > > For non-devs like myself who are interested in this project solely for the > machine learning aspect of it (which is my background), it would be great to > have a version of NuPIC that works out of the box. Something stable, even if > not optimized yet. > > In that case, maybe you should consider running a VM with preset NuPIC? > http://www.mail-archive.com/[email protected]/msg01012.html > If the download is slow, there's a torrent the guys are seeding, check that > out too. > > My interest in NuPIC, as I’d imagine is the case for most ML researchers, is > investigating its claims. How well it clusters? How well it captures temporal > data? Can I get better performance with it than state-of-the-art methods? > Most importantly, can I incorporate multiple layers in the hierarchy and > incorporate feedback? > > This is what I came to do, but ended up hacking everything aroud :) I would > be very interested to see your results. > > Best regards, Mark > > -- > Marek Otahal :o) > _______________________________________________ > nupic mailing list > [email protected] > http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
