Truth functions for composition rules
21 views
Skip to first unread message
Maxim Tarasov
Mar 22, 2022, 9:59:34 PM3/22/22
to open-nars
I am looking at some of the test cases in opennars, like this one and I'm wondering why the confidence scores don't seem to line up with what I would expect based on the Non-Axiomatic Logic book. I'm hoping someone can help clear this up.
As I understand it, truth value functions for composition rules all have confidence calculated as `c = and(c1, c2)` In the test case above, c1 is 0.90 and c2 is 0.80 so the resulting confidence should be 0.72 but the expected value instead is 0.81. Why is that?
Thanks!
As I understand it, truth value functions for composition rules all have confidence calculated as `c = and(c1, c2)` In the test case above, c1 is 0.90 and c2 is 0.80 so the resulting confidence should be 0.72 but the expected value instead is 0.81. Why is that?
Thanks!
Patrick Hammer
Mar 24, 2022, 9:35:06 AM3/24/22
to open-nars
Hi Max!
The confidence of the first premises is both 0.9 actually.
%0.80% means f=0.8 and c=0.9
%0.72;0.81% means f=0.72 and c=0.81
The notation is a bit confusing, which is also why in ONA I decided to print the truth value in an explicit way instead, mentioning both f and c.
Best regards,
Patrick
Maxim Tarasov
Mar 27, 2022, 7:51:54 PM3/27/22
to open-nars
Oh I see! That makes sense now. Thank you!
Maxim Tarasov
Mar 27, 2022, 7:57:02 PM3/27/22
to open-nars
I am working on a Swift implementation of the theory and for the life of me could not understand why the values were not lining up. I though I was going crazy there for a moment :) Thank you again, Patrick!
On Thursday, March 24, 2022 at 6:35:06 AM UTC-7 pat...@gmail.com wrote:
Reply all
Reply to author
Forward
0 new messages