If I'm recalling correctly from a numenta tutorial video, wasn't it that networks which went beyond the default number of columns (in terms of processing requirements) weren't deemed as efficient anyway? Seemed like the most efficient and appropriate methods to extract gains from running many networks (vs increasing column count) would be to build layers of networks - and more so the 'connective tissues' between them.
-- Michael Hale 219.448.0219 southshoremedia.com [email protected] <[email protected]> On Thu, Feb 5, 2015 at 12:11 AM, Tao Effect <[email protected]> wrote: > On this topic, I always felt from the first time I learned about NuPIC > that it would be perfect fit for Clojure. > > And now that I saw that a lot of work has been done on a Java > implementation of NuPIC, that possibility seems even more within reach. > > (Give it a consideration?) > > - Greg > > -- > Please do not email me anything that you are not comfortable also sharing with > the NSA. > > On Feb 4, 2015, at 2:56 PM, Matthew Taylor <[email protected]> wrote: > > Rich, > > We've been moving slowing in that direction, but there's a lot more > work to do. When Stewart made that comment, the NuPIC source code had > not yet even been open-sourced yet. Since then, we've made good > progress extracting the C++ code into the nupic.core repo, keeping the > python code in the nupic repo. > > The next step is in progress: > https://github.com/numenta/nupic.core/issues/314 > > Once that happens, nupic.core will be "algorithm-complete", and could > potentially be bound to different language implementations. > > Another task that progresses us towards this common goal is an > independent serialization format. Scott has been working on > implementing the Cap'n Proto serialization protocol within nupic.core. > You can track the progress on that initiative here: > https://github.com/numenta/nupic/issues/1449 > > We haven't even thought of a message passing protocol at this point... > so there is still lots of work to do but we're heading in that > direction. I don't think we'll be ready to start discussing message > passing until nupic.core has its own release schedule. > > You're the first person who's mentioned this topic in quite awhile. > Our current major initiative is getting NuPIC building and installing > on Windows, so that is where the effort is currently being spent. > > By the way, you can see our development roadmap here: > http://status.numenta.org/roadmap > > Regards, > --------- > Matt Taylor > OS Community Flag-Bearer > Numenta > > > On Wed, Feb 4, 2015 at 2:28 PM, Rich Morin <[email protected]> wrote: > > Back in 2013, Stewart Mackenzie started an interesting thread on > > Steps toward a distributed NuPic > > http://lists.numenta.org/pipermail/nupic_lists.numenta.org/2013-May/000090.html > > I've been having some similar ideas and wonder whether anything has > moved forward in this area. My implementation preference would be > Elixir/Erlang/OTP, but there are several other ways (eg, Clojure's > core.async, Go's Goroutines, Haskell) to dynamically produce message- > based suites of lightweight processes. > > Can anyone tell me of more recent work in this area? > > -r > > -- > http://www.cfcl.com/rdm Rich Morin [email protected] > http://www.cfcl.com/rdm/resume San Bruno, CA, USA +1 650-873-7841 > > Software system design, development, and documentation > > > > > >
