Could you explain more on Freq mode and non freq mode?

[xindi-dumbledore]

Hi :) I'm wondering what exactly are the freq mode and non-freq mode edge weight? Also, it seems to me non-freq is much slower than freq mode? (I always got stuck on "Building Cache") Thanks!

[xyjprc]

Hi @xindi-dumbledore , the Freq mode produces the count of sub-sequences, and the non freq mode is the normalized probability of subsequences.

For example, for observations
A -> B
A -> B
A -> C

Freq mode gives you
A -> B: 2
A -> C: 1

non-freq mode gives you
A -> B: 0.666667
A -> C: 0.333333

Note that this doesn't impact the higher-order rule extraction. Under the hood, the rule extraction works on the normalized probabilities. This freq mode and non-freq mode are just convenience functions for formatting outputs.

It is weird that freq and non-freq mode would behave differently. Can you reproduce the behavior, by closing python and re-running from scratch again for both methods? Let me know if you still keep seeing issues.