« first day (790 days earlier)      last day (1611 days later) » 

KronoS
KronoS
03:50
anybody around?
 
4 hours later…
Ivo Flipse
Ivo Flipse
07:38
@GreenGiant That sounds like a good idea, though I'd like to figure out when to apply multiple methods too. Because the robust/naive version works in about 70% of the time and it is supposed to work as real time as possible, so I'd prefer to keep the overhead down if possible
 
9 hours later…
GreenGiant
GreenGiant
16:18
In linear regression analysis, there is a value called "r squared" which will tell you how "confident" your algorithm was in determining its answer.
If this value is lower than some threshold, you could run additional algorithms.
Not sure if there's something similar for the kind of algorithm you have.

« first day (790 days earlier)      last day (1611 days later) »