pulsatrix 1.1.0
Loading...
Searching...
No Matches
global_sensitivity.hpp
Go to the documentation of this file.
1
5#pragma once
6
7#include <cstdint>
8#include <functional>
9#include <string>
10#include <vector>
11
12#include "pulsatrix/tensor.hpp"
13
14namespace pulsatrix {
15
24 std::vector<int64_t> features;
26 std::vector<float> lower;
27 std::vector<float> upper;
28};
29
32 std::vector<int64_t> features;
34 std::vector<float> mu;
36 std::vector<float> mu_star;
39 std::vector<float> sigma;
41 std::vector<float> mu_star_conf;
42 int64_t num_trajectories = 0;
43};
44
47 int64_t num_trajectories = 20;
50 int num_levels = 4;
51 int num_resamples = 1000;
52 float confidence = 0.95f;
53 uint64_t seed = 0;
54};
55
63[[nodiscard]] std::vector<std::vector<float>> MorrisSample(const GlobalSensitivityProblem& problem,
64 const MorrisOptions& options = {});
65
74[[nodiscard]] MorrisResult AnalyzeMorris(const std::vector<std::vector<float>>& samples,
75 const std::vector<float>& outputs, const MorrisOptions& options = {});
76
78[[nodiscard]] MorrisResult ComputeMorris(const std::function<Tensor(const Tensor&)>& predict,
79 const GlobalSensitivityProblem& problem, int64_t target_index,
80 const MorrisOptions& options = {});
81
85 std::vector<int64_t> features;
87 std::vector<float> first_order;
90 std::vector<float> total_order;
91 std::vector<float> first_order_conf;
92 std::vector<float> total_order_conf;
93 int64_t num_samples = 0;
94};
95
99 int64_t num_samples = 1024;
100 int num_resamples = 100;
101 float confidence = 0.95f;
102 uint64_t seed = 0;
103};
104
112[[nodiscard]] SobolResult AnalyzeSobol(const std::vector<float>& outputs, int64_t num_features,
113 const SobolOptions& options = {});
114
122[[nodiscard]] SobolResult ComputeSobol(const std::function<Tensor(const Tensor&)>& predict,
123 const GlobalSensitivityProblem& problem, int64_t target_index,
124 const SobolOptions& options = {});
125
126} // namespace pulsatrix
N-dimensional tensor. Owns its data buffer exclusively; a DeviceBackend* is injected (not owned) – th...
Definition tensor.hpp:29
Definition acquisition_functions.hpp:16
MorrisResult ComputeMorris(const std::function< Tensor(const Tensor &)> &predict, const GlobalSensitivityProblem &problem, int64_t target_index, const MorrisOptions &options={})
Samples, runs the model and analyzes: MorrisSample, then AnalyzeMorris.
MorrisResult AnalyzeMorris(const std::vector< std::vector< float > > &samples, const std::vector< float > &outputs, const MorrisOptions &options={})
Morris statistics from samples laid out as MorrisSample's (or SALib's) trajectories and one output pe...
SobolResult AnalyzeSobol(const std::vector< float > &outputs, int64_t num_features, const SobolOptions &options={})
Sobol indices from outputs in SALib's layout without second order: N blocks of (num_features + 2) out...
std::vector< std::vector< float > > MorrisSample(const GlobalSensitivityProblem &problem, const MorrisOptions &options={})
Morris trajectories (Morris 1991): each starts at a random grid point and moves one feature at a time...
SobolResult ComputeSobol(const std::function< Tensor(const Tensor &)> &predict, const GlobalSensitivityProblem &problem, int64_t target_index, const SobolOptions &options={})
Samples two independent uniform matrices A and B over the problem's ranges, builds the Saltelli desig...
The input space to explore: which features vary, over which ranges. Every other feature keeps its val...
Definition global_sensitivity.hpp:20
std::vector< float > upper
Definition global_sensitivity.hpp:27
std::vector< float > lower
Each varied feature's range, uniform between lower[i] and upper[i] (lower < upper).
Definition global_sensitivity.hpp:26
std::vector< int64_t > features
Flat indices of the features to vary; non-empty, no repeats.
Definition global_sensitivity.hpp:24
Tensor base
An instance shaped like the model's input.
Definition global_sensitivity.hpp:22
Morris settings.
Definition global_sensitivity.hpp:46
int num_resamples
Definition global_sensitivity.hpp:51
int num_levels
Grid levels per feature, even and >= 2; the step is levels / (2 (levels - 1)) of the range.
Definition global_sensitivity.hpp:50
float confidence
Definition global_sensitivity.hpp:52
int64_t num_trajectories
Definition global_sensitivity.hpp:47
uint64_t seed
Definition global_sensitivity.hpp:53
Morris screening results, one entry per varied feature (SALib's morris.analyze).
Definition global_sensitivity.hpp:31
std::vector< float > mu
Mean elementary effect: the average direction of the feature's effect.
Definition global_sensitivity.hpp:34
int64_t num_trajectories
Definition global_sensitivity.hpp:42
std::vector< float > sigma
Standard deviation of the elementary effects (ddof 1): large when the effect depends on where you are...
Definition global_sensitivity.hpp:39
std::vector< float > mu_star_conf
Half-width of the bootstrap confidence interval of mu_star.
Definition global_sensitivity.hpp:41
std::vector< int64_t > features
Definition global_sensitivity.hpp:32
std::vector< float > mu_star
Mean absolute elementary effect: overall importance (Campolongo et al. 2007).
Definition global_sensitivity.hpp:36
Sobol settings.
Definition global_sensitivity.hpp:97
int num_resamples
Definition global_sensitivity.hpp:100
float confidence
Definition global_sensitivity.hpp:101
uint64_t seed
Definition global_sensitivity.hpp:102
int64_t num_samples
Base samples N; the model runs N * (num_features + 2) times.
Definition global_sensitivity.hpp:99
Sobol indices, one entry per varied feature (SALib's sobol.analyze, first and total order).
Definition global_sensitivity.hpp:84
std::vector< int64_t > features
Definition global_sensitivity.hpp:85
std::vector< float > total_order
The share it explains including every interaction it takes part in. total - first is the feature's in...
Definition global_sensitivity.hpp:90
std::vector< float > first_order
The share of the output's variance the feature explains alone.
Definition global_sensitivity.hpp:87
std::vector< float > total_order_conf
Definition global_sensitivity.hpp:92
std::vector< float > first_order_conf
Definition global_sensitivity.hpp:91
int64_t num_samples
Definition global_sensitivity.hpp:93
N-dimensional tensor – owns a buffer via DeviceBackend*, RAII (Rule of Five).