Multiregression: objectives and metrics

Objectives and metrics

MultiRMSE

i=1Nd=1dim(ai,dti,d)2wii=1Nwi\sqrt{\displaystyle\frac{\sum\limits_{i=1}^{N}\sum\limits_{d=1}^{dim}(a_{i,d} - t_{i, d})^{2} w_{i}}{\sum\limits_{i=1}^{N} w_{i}}}

dimdim is the identifier of the dimension of the label.

Usage information

User-defined parameters

use_weights

Use object/group weights to calculate metrics if the specified value is true and set all weights to 1 regardless of the input data if the specified value is false.

Default: true

MultiRMSEWithMissingValues

d=1dimi=1NNum(ai,d,ti,d,wi)i=1NDen(ti,d,wi)\sqrt{\sum_{d=1}^{dim} \frac{\sum_{i=1}^N Num(a_{i,d}, t_{i,d}, w_i)}{\sum_{i=1}^N Den(t_{i,d}, w_i)}}

Num(a,t,w)={w(at)2, if tNaN0Num(a, t, w) = \begin{cases} w(a - t)^2, \space if \space t \neq NaN\\ 0 \end{cases}

Den(t,w)={w, if tNaN0Den(t, w) = \begin{cases} w, \space if \space t \neq NaN\\ 0 \end{cases}

dimdim is the identifier of the dimension of the label.

Usage information

User-defined parameters

use_weights

Use object/group weights to calculate metrics if the specified value is true and set all weights to 1 regardless of the input data if the specified value is false.

Default: true

Used for optimization

Name Optimization GPU Support
MultiRMSE + +
MultiRMSEWithMissingValues + -