Cheers for the feedback. Unfortunately Sequence Predictor is even worse (the SP owns that acronym), and a lot of people just blank when you try to explain how the TP is really a sequence memory.
I think the new Temporal Pooler is far enough out that we can come up with a new acronym for it, or possibly not even use the TP acronym for it at all (it's not going to be implemented as a single class as it involves at least two CLA layers). One possibility is to ditch the two-letter-acronyms and use a single word, such as "Predictor". We can continue to use the term TP as we do right now, and tell people it means Temporal Predictor, then gradually shift our shorthand to say Predictor alone. Changing the name in the code is the easy bit, but my experience is that everyone still uses the old acronym in their heads and will continue to do so for some time. I know it seems very pedantic, but naming is really important, especially when there can be such a large gap in shared understanding of how all these things fit together in NuPIC. So, the revised idea is: 1. Everyone says TP now means "Temporal Predictor" (we can start this right away) 2. Change "TP" and "Temporal Pooler" to "Predictor" and "Temporal Pooling" to "Temporal Prediction" in code, docs, white paper, etc. On Tue, May 13, 2014 at 12:26 AM, Chetan Surpur <[email protected]> wrote: > +1 Sequence Learner > On May 12, 2014 3:17 PM, "Subutai Ahmad" <[email protected]> wrote: > >> >> My first reaction was that it was a cool idea and certainly an accurate >> name. However, after reflection I agree that having two different things >> called "TP" would cause too much confusion. Nice thought though! >> >> Sequence Learner or Sequence Memory or Sequence Predictor are all good >> names. We have to port the TP to C++ as part of nupic.core so I suggest we >> call the class one of these names in that repo. At that point we can >> deprecate the existing TP10X implementation. >> >> --Subutai >> >> >> On Mon, May 12, 2014 at 1:34 PM, Matt Keith <[email protected]> wrote: >> >>> While changing to Temporal Predictor is an easy out for our predicament, >>> I feel that keeping the TP acronym for two different parts of the code will >>> continue to create confusion in the code and on the mailing list. I would >>> vote for officially changing the name to Sequence Memory and just change >>> all of the current TP references to SM to create a clean space for the new >>> TP implementation. I believe that a completely different name is the best >>> solution for the long term, and that should be our focus before we lock it >>> into the API. >>> >>> That's my 2 cents. Subutai and Scott, what is your opinion? >>> >>> Matt >>> >>> On May 12, 2014, at 2:07 PM, Matthew Taylor <[email protected]> wrote: >>> >>> > Not a bad idea. What do others think? I have no opinion, but looking >>> > at Subutai and Scott for their input... >>> > --------- >>> > Matt Taylor >>> > OS Community Flag-Bearer >>> > Numenta >>> > >>> > >>> > On Mon, May 12, 2014 at 1:03 PM, Fergal Byrne >>> > <[email protected]> wrote: >>> >> >>> >> In light of Jeff's new ideas about Temporal Pooling, every time we >>> mention >>> >> the current Temporal Pooler (TP) in NuPIC, we now have to use a >>> qualifier >>> >> like "the old Temporal Pooler" or "the thing we mistakenly called >>> Temporal >>> >> Pooler, but what we'll be now be calling Sequence Memory". >>> >> >>> >> Since the current TP in fact only does a single timestep prediction, >>> and >>> >> since everyone calls it the TP (in code, in docs, on the lists etc), >>> >> changing from "TP" to some other acronym will be slow, painful and >>> >> error-prone, my suggestion is to keep the initials but switch the >>> name to >>> >> "Temporal Prediction" or (if you want a more active name) "Temporal >>> >> Predictor". >>> >> >>> >> This name sheds anything about temporal pooling, and more correctly >>> >> describes the single thing it does - predict the SP(s) one step ahead >>> using >>> >> recent temporal context. >>> >> >>> >> If we can reach a consensus about this (or please provide a better >>> >> alternative!), we should implement it now in NuPIC before nupic.core >>> goes to >>> >> API release, and before Jeff's new Temporal Pooler begins to appear in >>> >> software. >>> >> >>> >> -- >>> >> >>> >> Fergal Byrne, Brenter IT >>> >> >>> >> Author, Real Machine Intelligence with Clortex and NuPIC >>> >> https://leanpub.com/realsmartmachines >>> >> >>> >> Speaking on Clortex and HTM/CLA at euroClojure Krakow, June 2014: >>> >> http://euroclojure.com/2014/ >>> >> and at LambdaJam Chicago, July 2014: http://www.lambdajam.com >>> >> >>> >> 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 >>> >> Join the quest for Machine Intelligence at http://numenta.org >>> >> 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 >> >> > _______________________________________________ > nupic mailing list > [email protected] > http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org > > -- Fergal Byrne, Brenter IT Author, Real Machine Intelligence with Clortex and NuPIC https://leanpub.com/realsmartmachines Speaking on Clortex and HTM/CLA at euroClojure Krakow, June 2014: http://euroclojure.com/2014/ and at LambdaJam Chicago, July 2014: http://www.lambdajam.com 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 Join the quest for Machine Intelligence at http://numenta.org Formerly of Adnet [email protected] http://www.adnet.ie
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
