Analysis results (Control dynamics)

This page summarizes insights gained from analysis.py which is for analyzing the inference control dynamics.

Concept use count histogram after example1.nal has been processed:

As we see, there are less concepts with higher use count than with low use count.

Priority of the concepts, again after example1.nal has been processed:

The pattern is even stronger for concept priority, which is expected as only a small amount of concepts should be in the attentional focus in the current moment, meaning high-priority concepts are the exception and allow the system to control its resources quite precisely.

Let's look at the average numbers:

Average priority: 5.664062500000024e-08
Average use count: 9.15203857421875

Assuming most concepts are very close to 0 in priority, it's also expected that the average value is close to 0. The average use count here is 9, but in any case has to be larger than 1, as a concept is used once at least when it is created.