David Ray

FYI, I already implemented this in HTM.Java and it works very well. If the `potentialRadius > -1` then the NAPI lets the configured value stand; if its -1, then it...

@natoromano I am concerned about this. Here's my reasons. The TM has a Connections local field coupling the particular Connections instance to a specific instance of a given TM. That...

> but the inputs are still reindexed I can't visualize this, but as long as there's a design for how this fits together, I'm cool with it. I just think...

> Well, let's say you are training a 32-cells TM with an external input which is another 32-cells TM. If the cell 5 from the external TM fires, it will...

Trying to tidy up my cognitive residue. This boils down to two parts: - The test should be changed so that it doesn't reference indexes that don't exist - although...

This is the gap I feel the Java version fills. It will be just as performant as the c++ version (for longer running processes due to cutting edge run time...

@houkx This has to be tested outside of Windows. I would suggest testing the Stream to see if it is null, and if it is, then in another try/catch, switching...

Actually the functionality mentioned in the title was not included in the PR linked to this issue so I'm reopening this as the link was in err.

Hi @willwandering, Does `grade test` also run the benchmarks? There is a flag which is documented on the project landing page [here](https://github.com/numenta/htm.java#does-my-code-work) which documents a method created to avoid running...

@rhyolight Will seems to have a lot of experience with this, and I'm in favor of it. I apologize @willwandering, I'm swamped right now and haven't had time to review...