HTM Implementers Meet N Greet - Yet Another HTM Implementation

[Deric_Pinto]

Hello HTM Hackers, I am also currently working on my own HTM Variant , I was wondering if everybody else is building there own representation , may be we can get together once in a month or so and sort of share ideas ,or at least brainstorm there own inspiration and paths everybody took to make it work. Like or please direct you msgs to me directly and I can set up something.

mailto : Deric.pinto@yahoo.com

Blockquote

[cezar_t]

Hi, what is different in your variant?

[jacobeverist]

Here is my variant: BrainBlocks

This one was focused on practical ML applications, speed and efficiency, and less focused on faithfulness to neuroscience accuracy.

My previous posts:

Releasing BrainBlocks 0.6: Building ML Applications with HTM-Like Algorithms

We are pleased to announce the public release of our software to the HTM community. Feel free to explore the README and the example code. We will be following up with more detailed analysis, discussion, and documentation of the more interesting and impactful features in the coming weeks. Below is a snippet of the README. Introduction BrainBlocks is a framework for building applications using principles from the theory of Hierarchical Temporal Memory. BrainBlocks focuses on solving general…

Releasing BrainBlocks 0.7.1: Building ML Applications with HTM-Like Algorithms

BrainBlocks 0.7.1 is released. We have previously posted about Brainblocks on this discussion board. This release includes a number of changes reflecting the user feedback provided by people here. We also included more comprehensive example scripts to see how each of the provided block libraries can be used. BrainBlocks is also provided as a pip package across all major platforms. Simple way to start is to install with pip: pip install brainblocks The current computational blocks pro…

[Deric_Pinto]

My variant has few key differentiations when it comes to HTM implementation:

1.Dendritic Computation :
My Implementation has the potential ability for the dendrites to do a lot more than just point dendrites in HTM where its just connected to another axon , it can do much more sophisticated computation and generate NMDA spikes based on those computations. ( 2021's Breakthroughs in Neuroscience and Other Biology - YouTube )
2.I don’t use a Spatial Pooler in my Implementation , I have a Temporal Pooler and an Apical Pooler. The Apical Pooler connects to the apical dendrites of the neurons from higher layers.
3. interCommunication between layers and Regions : In my implementation of HTM , I run a small 100* 100 block of neurons all stacked one on top of each other in sort of square shape , Which is a block a core computation Model , I can create multiple instance of the same block and they all can interconnect with each other through the temporal pooler and apical pooler based on situation where combined output needs to be generated for relevant input.

It could be argued that my implementation is not a HTM Implementation at all , As I have put in a whole bunch of changes I see fit and removed things I don’t see as needed.

Its an interesting project for me to work on for sure…

[cezar_t]

Sounds interesting, thanks. Unless there-s a special reason to carry this conversation in a private space, I hope no one will mind if it continues here

@Deric_Pinto

I can create multiple instance of the same block and they all can interconnect with each other through the temporal pooler and apical pooler based on situation

This means they “choose” to connect with each other dynamically or is just a thing you have to explicitly instruct them… statically (like defining a network topology from the start)

[Deric_Pinto]

The connections are driven based more dynamic associations, I do drive a strict Initial Network Topology to drive the learning process, but the nodes are independent and rely on randomness of the system and inputs given to them to connect with each other.