View Single Post
  #5   Spotlight this post!  
Unread 19-04-2014, 17:08
Ether's Avatar
Ether Ether is offline
systems engineer (retired)
no team
 
Join Date: Nov 2009
Rookie Year: 1969
Location: US
Posts: 8,034
Ether has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond reputeEther has a reputation beyond repute
Re: calling all Python gurus...

Quote:
Originally Posted by Spoam View Post
Just from a cursory glance, it looks like you're talking about implementing linear least squares, no?
No. L1 norm minimization.

Least Squares is L2 norm minimization, which is what is used for "OPR" computation here on CD. I am postulating that min L1 norm may be a better metric for FRC scouting purposes.

In the attached example, only 19% of the min_L2_norm computed alliance scores fall within +/-10 points of the actual alliance scores, whereas 41% of the min_L1_norm computed ones do.


Attached Thumbnails
Click image for larger version

Name:	L2 vs L1 residuals comparison.png
Views:	54
Size:	26.2 KB
ID:	16888  

Last edited by Ether : 19-04-2014 at 17:34. Reason: added graph
Reply With Quote