Extract patterns for analysis purpose


#1

Hey guys,

Do you have any technique for programmatically extracting patterns already learned by the HTM? I would like to perform an analysis on these patterns so to understand the data better. As a simple example if I have fed the following data sequence and the HTM learned that "c "comes after “b”, and at some point the HTM predicts c accurately,

abcd45ty8abcdjfipkabcd345869584abcd29840adfgkabcd

Programmatically, maybe using one of the HTM objects, how do I extract abc, assuming that the fact c comes after b cannot be easily detected by just manually scanning the data? Furthermore, is there such this as “forgetfulness” in HTM, such that we can configure how long the said patterns can be kept in memory?


#2

You would be well-informed to catch up on this thread.


#3

Thank you!