Quick Start
Get up and running with LOWESS in minutes.
Basic Smoothing
Smooth a noisy sine wave — the kind of signal where LOWESS shines. Each example recovers the underlying trend from 100 points of Gaussian noise.
#include <cmath>
#include <iostream>
#include <vector>
int main() {
const int n = 100;
std::vector<double> x(n), y(n);
for (int i = 0; i < n; ++i) {
x[i] = i * 2 * M_PI / (n - 1);
y[i] = std::sin(x[i]) + ((i * 7 + 3) % 17 / 17.0 - 0.5) * 0.6;
}
std::cout << "First smoothed value: " << result.y_vector()[0]
<< " (true: " << std::sin(x[0]) << ")\n";
return 0;
}
T & value() &
Definition fastlowess.hpp:150
Batch LOWESS model.
Definition fastlowess.hpp:804
Expected< LowessResult > fit(const std::vector< double > &x_values, const std::vector< double > &y_values, const std::vector< double > &custom_weights={})
Definition fastlowess.hpp:865
C++ wrapper for fastLowess library.
First smoothed value: 0.0272725 (true: 0)
With Confidence Intervals
#include <cmath>
#include <iostream>
#include <vector>
int main() {
const int n = 100;
std::vector<double> x(n), y(n);
for (int i = 0; i < n; ++i) {
x[i] = i * 2 * M_PI / (n - 1);
y[i] = std::sin(x[i]) + 0.1;
}
auto result = model.fit(x, y).
value();
auto lower = result.confidence_lower();
auto upper = result.confidence_upper();
double r2 = result.diagnostics().r_squared();
std::cout << "95% CI: [" << result.confidence_lower()[0] << ", " << result.confidence_upper()[0] << "]\n";
return 0;
}
double confidence
Definition fastlowess.hpp:91
double prediction
Definition fastlowess.hpp:92
Options for configuring LOWESS smoothing.
Definition fastlowess.hpp:194
IntervalsOptions intervals
Definition fastlowess.hpp:205
int iterations
Robustness iterations.
Definition fastlowess.hpp:196
double fraction
Smoothing fraction (0, 1].
Definition fastlowess.hpp:195
std::vector< std::string > outputs
Optional result components: diagnostics, residuals, weights, derivative, se, and sorted.
Definition fastlowess.hpp:211
95% CI: [0.292929, 0.372154]
Handling Outliers
LOWESS can robustly handle outliers through iterative reweighting:
#include <cmath>
#include <iostream>
#include <vector>
int main() {
std::vector<double> x_out = {1.0, 2.0, 3.0, 4.0, 5.0, 6.0};
std::vector<double> y_outlier = {2.0, 4.0, 6.0, 50.0, 10.0, 12.0};
auto result = model.fit(x_out, y_outlier).
value();
auto weights = result.robustness_weights();
for (size_t i = 0; i < weights.size(); ++i) {
if (weights[i] < 0.5) {
std::cout << "Point " << i << " is outlier (weight: " << weights[i] << ")\n";
}
}
return 0;
}
std::string robustness_method
Definition fastlowess.hpp:200
Point 3 is outlier (weight: 0)
Streaming Mode
For datasets too large to fit in memory, stream them in fixed-size chunks with overlap.
#include <cmath>
#include <iostream>
#include <vector>
int main() {
const int n = 5000;
std::vector<double> x(n), y(n);
for (int i = 0; i < n; ++i) {
x[i] = i * 10 * M_PI / (n - 1);
y[i] = std::sin(x[i] / M_PI) * std::exp(-x[i] / 30.0)
+ ((i * 7 + 3) % 17 / 17.0 - 0.5) * 0.3;
}
for (int start = 0; start <= 4000; start += 1000) {
int end = std::min(start + 1000, n);
model.process_chunk(
std::vector<double>(x.begin() + start, x.begin() + end),
std::vector<double>(y.begin() + start, y.begin() + end)
);
}
auto result = model.finalize().
value();
std::cout << "Smoothed " << result.y_vector().size() << " points in streaming mode\n";
return 0;
}
Streaming LOWESS model.
Definition fastlowess.hpp:907
Options for streaming LOWESS.
Definition fastlowess.hpp:235
int overlap
Negative (the default) means "use the library default" (chunk_size / 10, clamped to [1,...
Definition fastlowess.hpp:239
int chunk_size
Definition fastlowess.hpp:236
Smoothed 100 points in streaming mode
Next Steps