17

I'm potentially interested in using hierarchical temporal memory model to solve a research problem I am working on.

Are there any open source libraries for this? I'm fairly open to languages, although C++, Java or Haskell is preferred. If yes, has anyone had any experience with them?

nbro
  • 15,395
  • 32
  • 113
  • 196
shuttle87
  • 15,466
  • 11
  • 77
  • 106

8 Answers8

3

There's NuPIC (Numenta Platform for Intelligent Computing), which is now completely open-source. You also have NuPIC.Core (which contains the core NuPIC algorithms written in C++), but, at the moment, it is still under construction.

There's also one active implementation I could find on the Wikipedia page for the Memory-prediction framework (which is the J. Hawkins' theoretical framework for the HTM theory): Project Neocortex, which seems completely open source (researcher's page).

nbro
  • 15,395
  • 32
  • 113
  • 196
neural5torm
  • 773
  • 1
  • 9
  • 21
1

There is this Java HTM open source project: http://code.google.com/p/htm/

Paul Klemstine
  • 281
  • 1
  • 3
  • 9
0

pyHTM is a pure Python implementation.

nbro
  • 15,395
  • 32
  • 113
  • 196
Quonux
  • 2,975
  • 1
  • 24
  • 32
-1

I have created a simplified version, but it works at the moment; it's not that hard to use.

It's available here: http://devmaster.net/forums/topic/17960-feedback-working-out-of-heirarchical-temporal-memory/

It basically records the picture, then pumps it back out of the htm memory.

Jesse
  • 8,605
  • 7
  • 47
  • 57
Magnus Robert Carl Wootton
  • 1
  • 3
-1

You can now access to openSource HTM-CLA implementation. See here: http://blog.mohammadzadeh.info/index.php/open-source-implementation-of-hierarchical

softnhard
  • 14
  • 1
  • 1
-1

Since HTM was invented by the founders of Numenta, did you try their implementation? It's not technically open source, but it's free to do research with. I believe it's implemented in a mix of Python and C++.

Cerin
  • 60,957
  • 96
  • 316
  • 522
  • 1
    It's no longer available. I think numenta removed all their research documents as well as the source code that used to be in "legacy content". – Benjamin Crouzier Oct 27 '12 at 12:34
  • 1
    Yeah, I noticed this a couple months ago. It looks like Numenta was all hype and no substance. – Cerin Oct 27 '12 at 17:50
  • 1
    Well, I tink that numenta has developed a lot of useful algorithms (see vitamindinc for example, which is based on numenta algorithms). They took it out apparently because nothing great was coming from the open source community. – Benjamin Crouzier Oct 27 '12 at 17:53
  • 1
    Numenta's HTM algorithms are entirely open-source. Even their research algorithms are open source. See numenta.org. – Matthew Taylor Oct 31 '14 at 17:30
  • Numenta's NuPIC HTM code is CPython + SWIG + C++, and was available under a dual license. There is now a community fork of NuPIC called htm.core, which is AGPL-only. – dstromberg Aug 20 '19 at 15:18
-1

Another open-source codebase here:

http://code.google.com/p/adaptive-memory-prediction-framework/

This one is an implementation of MPF based on Kohonen SOM, not HTM.

user1191311
  • 85
  • 1
  • 1
-2

Yeah Numenta bailed and decided to go for the bucks themselves but it looks like some of guys on the Forums are trying to get an open source version going. Try looking here:

http://sourceforge.net/p/openhtm/wiki/Home/

bobo
  • 1
  • 1
    Numenta was the company set up by Jeff Hawkins to continue the research and build the software. In June 2013 they open-sourced the core software, NuPIC, with a growing community around it. OpenHTM have joined us recently. See [Numenta.org](http://numenta.org) to get involved. – Fergal Byrne Nov 12 '13 at 17:18