|
fastlowess-cpp
High-performance LOWESS smoothing
|
Options for streaming LOWESS. More...
#include <fastlowess.hpp>
Public Attributes | |
| int | chunk_size = detail::k_default_chunk_size |
| int | overlap = detail::k_default_overlap |
| Negative (the default) means "use the library default" (chunk_size / 10, clamped to [1, chunk_size - 10]). | |
| std::string | merge_strategy |
| weighted_average, average, take_first, take_last | |
Public Attributes inherited from fastlowess::LowessOptions | |
| double | fraction = detail::k_default_fraction |
| Smoothing fraction (0, 1]. | |
| int | iterations = 3 |
| Robustness iterations. | |
| double | delta = NAN |
| Interpolation threshold (NaN = auto) | |
| std::string | weight_function = "tricube" |
| std::string | robustness_method = "bisquare" |
| std::string | scaling_method = "mad" |
| std::string | boundary_policy = "extend" |
| std::string | zero_weight_fallback = "use_local_mean" |
| IntervalsOptions | intervals |
| std::optional< uint64_t > | seed |
| double | auto_converge = NAN |
| Auto-convergence threshold. | |
| std::vector< std::string > | outputs |
| Optional result components: diagnostics, residuals, weights, derivative, se, and sorted. | |
| bool | parallel = true |
| std::string | backend = "cpu" |
Execution backend: "cpu" (default) or "gpu" (requires the library to be built with the gpu Cargo feature and a Vulkan/Metal/DX12-capable GPU). | |
| std::string | missing = "error" |
| Policy for non-finite (NaN/Inf) values in input data: "error" (default) or "drop". | |
| CVOptions | cv |
| Cross-validation configuration. Empty fractions disable CV. | |
| std::vector< double > | custom_weights |
| Per-observation case weights. | |
| bool | retain_model = false |
Retain the fitted model's training data, enabling LowessResult::predict_model(). | |
Options for streaming LOWESS.
| int fastlowess::StreamingOptions::chunk_size = detail::k_default_chunk_size |
| std::string fastlowess::StreamingOptions::merge_strategy |
weighted_average, average, take_first, take_last
| int fastlowess::StreamingOptions::overlap = detail::k_default_overlap |
Negative (the default) means "use the library default" (chunk_size / 10, clamped to [1, chunk_size - 10]).