Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign upAdding MAE as a standard output for regression tasks #657
Conversation
|
Can you add a note on this to |
|
There is no section for the next/next release and I am unaware of the version numbering you would like to use. Would you like me to create a section for |
yes.
yes, thanks |
|
No problem. Suggestions implemented. |
|
Thanks! |
|
Hello. |
|
Go ahead and file a separate issue and carefully read the instructions on the issue package to create a reproducible example. Also, in the issue, please define MBE and NSE (nonstandard evaluation?) |
Regression tasks out-of-the-box are showing only RMSE and R^2; MAE has been added. Given that a lot of algorithms working with L1 norms I think it will be beneficial for users to immediately be able to check absolute errors too. L1 is not some arcane specialised norm and I think adding MAE to the defaults will save valuable time (and potential bugs) for people writing their own cost functions to add MAE-like measures.