Hi guys,
Is there a general guide out there for developing SDR encoders? Or perhaps a test suite for testing valid/acceptable generated SDRs?
Cheers
Hi guys,
Is there a general guide out there for developing SDR encoders? Or perhaps a test suite for testing valid/acceptable generated SDRs?
Cheers
There is no test suite. You’ll have to figure out if the semantics work out by testing it yourself. Sometimes I compare the minicolumn activations the SP creates to previously seen output for the similar input and make sure there is a high overlap. That at least tells you that the SP is recognizing your semantics. You can see an example of that here.
Also, for resources, be sure to search the forum. Lots of stuff about encoders here.
Also, there is this document called Building HTM Systems I’m actively working on, which has a section on encoders.
Thanks for this, I am keen to use SDR’s but have struggled finding time to really focus on the subject matter and get a use case of my own working.
Can I offer my assistance with your Building HTM Systems documentation?
I am a solutions architect and coder so could probably assist with documentation, proof reading and code examples.
Thank you very much for the response and the references. These are very helpful. My goal and as an exercise is to try write an encoder for an image data or perhaps sound data. I’ll start reading on the documentation you are building.
Some further resources for you:
Read: Encoding vision for HTM
There are some links at HTM School under “More Encoder Resources”, if you haven’t found them already.
@rhyolight and @dub3 Cheers