tlby commented on issue #14874: [MXNET-1399] multiclass-mcc metric enhancements
URL: https://github.com/apache/incubator-mxnet/pull/14874#issuecomment-489397088
 
 
   I don't think it's absolutely necessary, but I definitely regret the name 
choice because while the implementation came from a discrete derivation of the 
Pearson correlation coefficient, all a user cares about is that it's a 
multiclass extension of Matthews correlation coefficient.  When it comes to 
using these metrics Pearson is for continuous and Matthews is for discrete 
predictions.  This metric is discrete so I feel it was misnamed.  I later 
discovered that the formula I used is equivalent to one [mentioned in 
Wikipedia](https://en.wikipedia.org/wiki/Matthews_correlation_coefficient#Multiclass_case)
 as an extension of Matthews.
   
   Perhaps we should leave "pcc" as an alias for the sake of reverse 
compatibility?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to