|
pulsatrix 1.1.0
|
First-order accumulated local effects (Apley and Zhu 2020) of one feature. More...
#include <ice.hpp>
Public Attributes | |
| int64_t | feature_index = 0 |
| int64_t | target_index = 0 |
| std::vector< float > | edges |
| Bin edges, strictly increasing: the minimum, then the upper edge of each bin. | |
| std::vector< float > | effects |
| The centered accumulated effect at each edge. | |
| std::vector< int64_t > | counts |
| Instances in each bin (edges.size() - 1 of them). | |
| std::vector< float > | feature_values |
| Each instance's own value of the feature, for a rug. | |
First-order accumulated local effects (Apley and Zhu 2020) of one feature.
The feature's range is cut at quantiles into bins holding about equal numbers of instances. Within each bin, every instance in it is moved to the bin's lower and upper edge and the change in the output is averaged; the averages are summed from the left and centered so their count-weighted mean is zero. Because each instance only moves within its own bin, ALE never reads the model at combinations of feature values the data doesn't contain, which is where PDP misleads when features are correlated.
| std::vector<int64_t> pulsatrix::AleResult::counts |
Instances in each bin (edges.size() - 1 of them).
| std::vector<float> pulsatrix::AleResult::edges |
Bin edges, strictly increasing: the minimum, then the upper edge of each bin.
| std::vector<float> pulsatrix::AleResult::effects |
The centered accumulated effect at each edge.
| int64_t pulsatrix::AleResult::feature_index = 0 |
| std::vector<float> pulsatrix::AleResult::feature_values |
Each instance's own value of the feature, for a rug.
| int64_t pulsatrix::AleResult::target_index = 0 |