HTM, and licensing

I would like to apply the HTM theory in some competitions. I would not use NuPIC, nor other related software developed by NuPIC community. Just my own implementations.
Competition at which has a money prize usually demands use of the winning software for commercial purpose (license).
I have read Numenta License Guide. I figured out that unfortunately there is no point to use HTM theory at kaggle because competition founder would have to apply for IP license to Numenta. In other words a submission will be ineligible to win a prize.

Do I understand it correctly?

1 Like

Not exactly, but it’s a tricky subject. I asked some similar questions two years ago on the Kaggle forums. The clarification I got from the Kaggle admin was this:

A Submission will be ineligible to win a prize if it was developed using code containing or depending on software licensed under an open source license:

  • other than an Open Source Initiative-approved license (see; or
  • an open source license that prohibits commercial use.

AGPL is OSI-approved, and allows commercial use (just not proprietary commercial use). So you should be ok.


IANAL, and don’t work for Numenta, but have discussed this with people there. This is my understanding.

If you’re not using NuPIC, then this depends on a) what license you use for your implementation (AGPL only applies to NuPIC itself), and b) what potentially Numenta-owned IP you incorporate in your implementation.

This is a grey area, because Numenta has only explicitly placed their patents and IP into NuPIC (and possibly now, not into anything developed independently. Their informal attitude (ie this is not their legal declaration) is that a project which uses the AGPL or equivalently strong license, and refers to Numenta for commercial licensing, will not have a problem.

One thing is for sure. If you use any Numenta-owned IP for your project and do not open source it, then you’ll have to talk to Numenta about what that implies.

1 Like

Let me try to clarify.

As Matt said, it is clear that if you use NuPIC, you do so under the AGPL, and this seems consistent with the Kaggle rules, so there is no problem there.

However, if you do your own implementation, as Fergal said, then the situation is a bit murkier. Of course, if this implementation does not read on our patents, there’s no problem. However, if it does rely on our intellectual property, then the following applies. In the event you are using our IP for non-commercial purposes, then we have made a clear non-assert statement. So if you do not plan to commercialize your Kaggle entry, then it is fine with us. However, I’m not sure how that squares with the Kaggle rules, so you have to evaluate that yourself - but any issue would be on the Kaggle side, not the Numenta side.

If you do want to commercialize your entry, or it is required by Kaggle that you are able to do so, you would need an IP license from us, which we would be willing to discuss. I like the idea of Kaggle competitors using our technology, so we’d figure out a way to give you an appropriate license.