Hi, I'm wondering if this premise is true:
An untrained spatial pooler should create randomly distributed SDRs. This > is way too structured. My understanding of the spatial pooler is that it "learns" to select columns to represent a *semantic* space where given all the semantic possibilities in a particular problem domain, all the columns of the spatial pooler will represent each possible input of a that problem domain. So if input bits are clustered together its because those bits represent the meaning (semantics) represented in that input? I always thought that as long as the SDR produced by the SpatialPooler was sparse (~2%), and similar inputs have a proportional number of columns in common (who's proportionality is determined by the "sameness" of the input), then it didn't really matter **which** columns were chosen to select the input, or whether for certain inputs they tended to be clustered at certain locations? This is interesting. Could someone please clarify what about my understanding is true/false? :-) Cheers, David On Fri, Apr 1, 2016 at 2:53 AM, Marcus Lewis <[email protected]> wrote: > Hi all, > > Here's a quick post: *The column SDR that wasn't random enough* > <http://mrcslws.com/blocks/2016/03/31/column-sdr-not-random-enough.html> > > I noticed an issue with my spatial pooler, and I couldn't think of why it > was happening. This blog post is my debugging notes. :) > > Marcus > -- *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
