Subutai said that the ScalarEncoder should do better than the AdaptiveScalarEncoder. And the RDSE will probably be the same as the Scalar. Apparently, there are some manual things that need to be done in order to allow the swarm process to include the RDSE in the pso parameters.
--------- Matt Taylor OS Community Flag-Bearer Numenta On Mon, Mar 24, 2014 at 3:02 PM, Chetan Surpur <[email protected]> wrote: > My guess is that since the sine wave has a fixed range, the ScalarEncoder > would work as well or better than the AdaptiveScalarEncoder. > > As for the RDSE, I believe you also need to specify the 'resolution' > parameter (also, a couple of the parameters you specified are not > necessary). > > > On Mon, Mar 24, 2014 at 2:56 PM, Matthew Taylor <[email protected]> wrote: > >> Yes, the swarm picked a ScalarEncoder, but I wonder why it didn't pick >> the AdaptiveScalarEncoder or the RDSE. I'm not very familiar with swarms, >> so I'm not sure. When I tried to replace the "ScalarEncoder" in the >> model_params.py file with "RandomDistributedScalarEncoder", I got this >> message: >> >> Error in constructing RandomDistributedScalarEncoder encoder. Possibly >> missing some required constructor parameters. Parameters that were provided >> are: {'maxval': 1.0, 'name': 'sine', 'minval': -1.0, 'clipInput': True, >> 'n': 427, 'w': 21} >> >> It worked ok when I used "AdaptiveScalarEncoder", however, so this makes >> me wonder if the swarming process will include the RDSE as a possibility or >> not? >> >> --------- >> Matt Taylor >> OS Community Flag-Bearer >> Numenta >> >> >> On Mon, Mar 24, 2014 at 2:01 PM, Fergal Byrne < >> [email protected]> wrote: >> >>> Great question Hideaki. I have a feeling it's using the old scalar >>> encoder, which only has a small number of SDR's. It'd be interesting to >>> re-run that with the RDSE set to a finer resolution (Chetan?). >>> >>> Regards >>> Fergal Byrne >>> >>> >>> On Mon, Mar 24, 2014 at 8:42 PM, Hideaki Suzuki <[email protected]>wrote: >>> >>>> Thank you Matt, >>>> This is a really cool tutorial! >>>> >>>> I was wondering what was causing prediction gaps and what can bring >>>> better precision... turning off/down SP boosting? Anyway, that is a >>>> different issue. >>>> >>>> The screencast is definitely easy to understand for me. Nice. :) >>>> >>>> >>>> >>>> 2014-03-21 11:41 GMT+09:00 Matthew Taylor <[email protected]>: >>>> >>>>> Subutai and I have been working on some examples of using NuPIC to >>>>> predict sine waves, which is a common task NuPIC newbies attempt when >>>>> experimenting. >>>>> >>>>> He has put together an example that walks through several examples of >>>>> swarming with sine wave data step by step: >>>>> >>>>> https://github.com/subutai/nupic.subutai/tree/master/swarm_examples >>>>> >>>>> And I have created a screencast that goes over programmatic swarming >>>>> and model creating using sine data: >>>>> >>>>> http://www.youtube.com/watch?v=KuFfm3ncEwI >>>>> >>>>> I hope this helps some of you trying to figure out different ways of >>>>> using NuPIC. Enjoy! >>>>> >>>>> --------- >>>>> Matt Taylor >>>>> OS Community Flag-Bearer >>>>> Numenta >>>>> >>>>> _______________________________________________ >>>>> 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 >>>> >>>> >>> >>> >>> -- >>> >>> Fergal Byrne, Brenter IT >>> >>> <http://www.examsupport.ie>http://inbits.com - Better Living through >>> Thoughtful Technology >>> http://ie.linkedin.com/in/fergbyrne/ >>> https://github.com/fergalbyrne >>> >>> e:[email protected] t:+353 83 4214179 >>> Formerly of Adnet [email protected] http://www.adnet.ie >>> >>> _______________________________________________ >>> 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 >> >> > > _______________________________________________ > 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
