Hi Mark and Aaron, and welcome to the NuPic community. I few tips/pointers:
- hierarchy: we are not doing this yet, NuPic is working with one region for now. - sensible defaults and components: https://github.com/numenta/nupic/wiki/Running-Swarms will explain some of this. If you use the tool-set that we call OPF (Online Prediction Framework) you will be able to run a swarm that should optimize for you. - music analysis: a couple of things you might be interested in. And impressive first hackathon demo was note prediction for a song. This was based on midi sequence of notes for training. Timing was not predicted or accounted for, only the sequence, but it was still impressive. - audio analysis: converting raw audio waveform via FFT into a spectrum and convert the frequency scalar values. Your in luck - Stewart just posted yesterday on this list a simple python file that does this I think. I find a large portion of software engineers have a fascination with music and sound recording and production, myself included. I expect there will be quite a bit of interest in music, speech and audio applications of NuPic. -Doug On Sun, Nov 24, 2013 at 7:59 AM, Mark Wilson <[email protected]> wrote: > I'm also a lurker and also interested in signal processing (not music, but > probably near enough). > > I'm a bit overwhelmed by lots of information about what all the different > parts of the Nupic system do. I have read the CLA Whitepaper and have a bit > of a grasp on the concepts. What I am having difficulty with is bridging > the gap between that and the signal I would like to infer and predict. > > Can some one help outline what steps should be taken to set up a > "suitable" HTM? Things I'm thinking about: > > - HTM regions and hierarchy > - Which Nupic libraries should be used (OPF, more direct CLA > libraries?) > - Some sensible defaults for the multitude of tunables. > - Maybe there is another project that is similar enough to be used? > > Any help or suggestions would be great. > > Thanks > > Mark > > > On 14 November 2013 14:07, Aaron Zimmerman <[email protected]>wrote: > >> Hi all, >> >> I’ve been a lurker for a while, excited to try my hand at some CLA >> applications, especially toward music analysis and creation. >> >> Is there any documentation of the different types of inferences and the >> requirements to use them? >> >> The hot gym example uses “TemporalMultiStep”. While I can infer some >> meaning from the name, it isn’t clear to me what the technical difference >> is between this and other. >> >> >> Thanks, >> >> Aaron Zimmerman >> Platform Engineer >> Sprout Social >> 773.227.7528 >> @apzimmerman <https://twitter.com/apzimmerman> >> sproutsocial.com<http://www.google.com/url?q=http%3A%2F%2Fsproutsocial.com%2F&sa=D&sntz=1&usg=AFrqEzcFZhOCY3atDfnGIaMM9Kza3cB4Zw> >> >> _______________________________________________ >> 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
