We’re proud to announce the first release of the “HTM.Core”, the HTM-Community’s successor to the NuPIC.core library. Although this library started as a fork of Numenta’s NuPIC.core, we’ve changed many things and so we’ve also changed the name to “HTM.Core”. We have made a large number of improvements to NuPIC with the goals of making it easier to use and faster to run. For instructions on how to use this software, visit our github page: https://github.com/htm-community/htm.core
This is a collaborative effort. If you have a question, encounter a problem, or have an idea for how to improve this software, then please tell us about it on our github page or on this forum.
Thank you,
Breznak, Dkeeney, Dmac, and everyone else who contributed.
haven’t been around in a while… trying to catch up. Could anyone help bring me up to speed?
This project on github “htm.core” is the new authoritative code repo implementing numenta’s ideas?
It seems like the idea was to remove stewardship of the code from numenta’s hands and to the OSS community?
The goal is not to maintain backward compatibility with the old nupic core api? Therefore the existing language bindings (java/js/etc) will need updated/ported to htm.core?
numenta is fully supportive of this change of repo/api/ownership?
Thanks, just want to make sure I understand the landscape, hoping to dig back into this exciting space.
There are many HTM codebases implementing Numenta ideas. I have been promoting htm.core as a good place for folks wanting a Python 3 implementation to work from, since NuPIC will not be updated from Python 2.
Correct, there is no real reason for backwards compatibility. Htm.core will not be a drop-in replacement for NuPIC / nupic.core.
There are no current language bindings to nupic.core aside from NuPIC (python 2). All other HTM implementations I know about are not bindings to nupic.core, but stand-alone libraries implementing HTM independently.
There really is no explicit change of ownership. US copyright laws states that the author of the code owns the copyright to the code. In the case of htm.core, as the code changes gradually over time, less Numenta code exists while more is contributed by the community. This can happen with any open source codebase. It can be forked into another project and taken in other directions. We always knew (and hoped!) this would happen when we open sourced the codebase. We want these ideas to live on. They are more important than the code implementations.
I am fully supportive of the htm.core codebase. I have encouraged everyone involved in this project at every step of the way, with full visibility of the rest of the Numenta engineering team. We all wish them the best as we focus on our current work integrating brain-inspired ideas into current deep machine learning networks.
@rhyolight thanks for the solid endorsement for htm.core.
This is the repository that was formerly call nupic.cpp. Numenta asked us to rename it (and all namespaces) to htm.core to emphasize that this is a community effort.
Although the API has changed a little from NuPIC the underlining concepts are still based on the Numenta’s algorithms. The intent was to modernize the code (both C++ and Python) and to improve its performance.
If anyone has a problem re-targeting any code previously using the NuPIC libraries, please let us know in the HTM.core Issues forum and we will try to help you out. It is not quite a plug-in replacement but it normally does not take very much tweaking to get it running.