Developer Guide for Intel® Data Analytics Acceleration Library 2019 Update 5
The metrics are computed given the input data meets the following requirements:
is non-zero. Returns error otherwise.
The PCA algorithm receives input argument eigenvalues
. It represents the following quality metrics:
The library uses the following quality metrics:
|
Quality Metric |
Definition |
|---|---|
|
Explained variance |
![]() |
|
Explained variance ratios |
![]() |
|
Noise variance |
![]() p r - number of principal components, p - number of features in the data set |