Hi Aseem, yes, temporal prediction is a core functionality of the CLA. You can ask it for prediction of a sequence n-steps head, or do it in a simple loop (eg to predict next 5 steps):
for (step=1, step<=5, step++): step = CLA.predictNext() print step CLA.feed(step) We have some toy-code that does just what you describe - it learns a temporal sequence, let's the user define first few input steps and then continues - can be found at https://github.com/chetan51/linguist Cheers, breznak On Wed, Jan 8, 2014 at 12:08 AM, Matthew Taylor <[email protected]> wrote: > Hi Aseem, > > I think you're asking if you can get a classification of the current > sequence(s) from NuPIC instead of only predictions and anomalies. I'm > not sure what the answer is. The TP does sequence memory, so the > current sequences for the current state should be stored within the > CLA's state somehow, but I don't know how this might be exposed > through the NuPIC API. I would have to ask Scott or Subutai (CCed). > > Thanks, > --------- > Matt Taylor > OS Community Flag-Bearer > Numenta > > > On Thu, Jan 2, 2014 at 10:59 AM, Aseem Hegshetye <[email protected]> > wrote: > > Hi, > > Its awesome that Matt taylor is full time Nupic member now!! > > > > I am working on building CLA for sequence prediction. My current version > learns a sequence and then predicts it as it is. > > I havent used look up tables, and memory is stored in connections > between cells. > > I plan on making it learn multiple sequences and then as user provides > first few inputs of a particular sequence, it should predict the remaining > sequence. I plan on designing 2nd and 3rd order connections. > > where can this sequence detection be used? I believe it can be used in > robotics, but would it also be used in data mining. > > Also if motor behaviour gets linked up with CLA ( memory) this sequence > detection would have some applications i guess. > > > > I haven't used the CLA that is in nupic repository, but i know it > predicts anomalies and nth step prediction. I want to know if it predicts > sequences. To make it predict a whole sequence do we have to keep n number > of look up tables? > > > > Regards !! > > Aseem Hegshetye > > > > _______________________________________________ > > 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 > -- Marek Otahal :o)
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
