Hi, all those interested.
A quick update.
Basically for this project at this time, my first goals have now been met.
In the end I did not modify the HTM like project. I implemented my own project from scratch.
There is still a lot which could be done to improve the engineering. But as a goal I have been able to test some ideas for the core algorithm, which is what I wanted to do.
I’m now thinking most about the core algorithm.
For that, as I have said, I am not going to go public with details at this time.
In general terms what I want to do the same. As I stated in the original presentation:
'I want to “tune” the network using feedback, dampened by inhibition, to isolate resonances associated with different sub-networks in the connections.
In terms of sub-networks, the output should be similar to networks identified from MRI data using recursive modularity analysis. Say, like those from this paper:
Also in my “point 4” in another list:
‘To be brief those general principles are that I think the important structure for perception and prediction will be hierarchies of sub-networks in the matrix of adjacency relations between observations of the world (currently just looking at texts and seeking language phrase structure.) Different inputs (each new sentence) will highlight different such hierarchies of sub-networks. The key task I see is to identify hierarchical sub-networks expanding away from each new input.
Most simply: to find hierarchical sub-networks in the adjacency matrix we build.’
If anyone has some good ideas how to do that, by all means post them here and let’s discuss them.
This is the problem I am thinking about at the moment.