Sitting in NuPIC issues here; https://github.com/numenta/nupic/issues/2483 An equivalent issue doesn't look like it's been raised in core though. I know I was awaiting/watching movement in NuPIC to tackle this in the Core.
On Wed, Oct 14, 2015 at 5:43 PM, cogmission (David Ray) < [email protected]> wrote: > Yes, I know we talked about removing empty synapses which attribute to the > overall synapse count, but I can't tell you whether it is in github or not > and if it is slated to be worked on? I think I did implement it in HTM.java > though... > > On Wed, Oct 14, 2015 at 11:23 AM, Richard Crowder <[email protected]> > wrote: > >> Chetan increased the ranges a while back (June) in this commit >> https://github.com/numenta/nupic.core/commit/54c1ab975f784971637bf866582de9f49e8c0f87#diff-dff563d2b4cacf6184c85c53ed96cf75 >> The defaults could be increased. If you need the indicies types changing to >> larger types, we may need to look at the serialization (?). Worth trying to >> change those locally Jos. I can help out if needed. >> >> nupic.audio is currently (stalled due to my Windows porting work) >> intended to use core only btw. The genre classification being the most >> fleshed out so far. But not using SP & TM in earnest yet. >> >> David, I do believe an issue was raised to cull synapse empty segments. >> >> Regards, Richard. >> >> On Wed, Oct 14, 2015 at 5:21 PM, cogmission (David Ray) < >> [email protected]> wrote: >> >>> RE: Orphan Synapses - Did we create the issue and implement it - that >>> deals with segment removal once all synapses have been culled? I mentioned >>> this before and now it seems we have a real-life customer who bumped up >>> against this issue? >>> >>> On Wed, Oct 14, 2015 at 11:08 AM, Matthew Taylor <[email protected]> >>> wrote: >>> >>>> Jos, is your code available? If not, can you push it to github? I would >>>> love to see how you are utilizing the C++ only parts of NuPIC (nupic.core). >>>> As far as I know, no one else is doing this right now (correct me if I am >>>> wrong, community!) >>>> >>>> Regards, >>>> >>>> --------- >>>> Matt Taylor >>>> OS Community Flag-Bearer >>>> Numenta >>>> >>>> On Wed, Oct 14, 2015 at 8:53 AM, Jos Theelen <[email protected]> wrote: >>>> >>>>> My temporal memory (or better the connections in it) throws an >>>>> exception: cannot create synapse: segment has reached maximum number of >>>>> synapses. I looked around if anybody else has had this problem, but found >>>>> nothing. I work with a 32 bit C++ program, made from the nupic.core >>>>> sources. The parameters I used are the same as for the hotgym-example, >>>>> that >>>>> gave no problems. >>>>> >>>>> I am now trying different parameters for the temporal memory, but I >>>>> have no idea which ones I should change. Can somebody say what could be >>>>> the >>>>> reason for this exception? >>>>> >>>>> >>>> >>> >>> >>> -- >>> *With kind regards,* >>> >>> David Ray >>> Java Solutions Architect >>> >>> *Cortical.io <http://cortical.io/>* >>> Sponsor of: HTM.java <https://github.com/numenta/htm.java> >>> >>> [email protected] >>> http://cortical.io >>> >> >> > > > -- > *With kind regards,* > > David Ray > Java Solutions Architect > > *Cortical.io <http://cortical.io/>* > Sponsor of: HTM.java <https://github.com/numenta/htm.java> > > [email protected] > http://cortical.io >
