Skip to content
fastloess

Changelog

Changelog

Section titled “Changelog”

This changelog includes end-user changes only. For internal development notes, see the repository changelog.

3.0.0

Section titled “3.0.0”

Added

Section titled “Added”
  • Added OnlineLoess.add_point_vector() for multivariate Online point updates.
  • Added a version export at fastloess/version so consumers can query the Node.js package version without loading the native addon.
  • Added cv to Batch options for grouped cross-validation configuration alongside legacy CV fields.
  • Added outputs arrays to Batch, Streaming, Online, and prediction options for grouped optional result selection.
  • Added retain_model and LoessResult.predict(newX, options) for out-of-sample prediction.
  • Added return_gradient to SmoothOptions, StreamingOptions, and OnlineOptions.
  • Added confidence_intervals/prediction_intervals/return_se to StreamingSmoothOptions and OnlineSmoothOptions. Online requires update_mode: "full" or throws. New OnlineOutput bound fields.

Changed

Section titled “Changed”
  • Clarified Batch residual_sd as 1.4826 * MAD; Streaming reports the cumulative sample standard deviation of emitted residuals.
  • Breaking change: replaced individual return_* output booleans with outputs: [...] for Batch, Streaming, Online, and prediction options.
  • Breaking change: replaced flat interval options and prediction levels with intervals: { confidence, prediction }; CV uses only cv: { fractions, method, k } with an outer seed, replacing flat CV options and cv.seed.

Fixed

Section titled “Fixed”
  • Use the fitted model’s kernel for unweighted, unpadded one-dimensional linear prediction standard errors without robustness downweighting, with smoothing influence rows and global residual degrees of freedom, including interpolation; align prediction-interval residual scales.
  • Correct tied-coordinate interpolation splits and normalized cell geometry, add two-dimensional neighboring-edge blending, and match R’s neighborhood rounding near integer span boundaries.
  • Preserve zero interpolation subdivision thresholds to match R’s small-span cell trees, retaining terminal observation vertices and correcting interpolated fits on small datasets.
  • Normalize local case weights by their neighborhood maximum so common scaling cannot turn valid positive weights into an epsilon-triggered unweighted fallback.
  • Prevented overflow in even medians, mean/bisquare scales, Batch/Streaming diagnostics and AIC, and local/all-tied weight sums for large finite inputs.
  • Keep generated native version checks tied to package.json, and provide TypeScript declarations for the fastloess/version subpath.
  • Reject unknown output names across Batch, Streaming, Online, and prediction options. Preserve the fit_async() TypeScript return type as Promise<LoessResult> after N-API builds.
  • Validate case-weight lengths and values before dropping missing observations, so invalid weights on dropped rows are not silently ignored.
  • Preserve case weights through sorted CV training subsets and multidimensional predictions. Serial and parallel CV now agree on seeded folds and held-out LOOCV predictions; K-fold counts above the retained observation count are rejected.
  • Reject non-positive or non-finite Streaming/Online auto-convergence tolerances. Online auto-convergence requires full updates with robustness iterations.
  • Include all observations for Gaussian smoothing and prediction while preserving the k-th-neighbor bandwidth; use the true exponential without an artificial tail floor.
  • Correct case-weighted standard errors for direct, unpadded one-dimensional linear fits and retained prediction, with matching serial/parallel local moments. Span-one fits no longer substitute a kernel-free global OLS formula.
  • Honor configured zero-weight fallback policies in constant-degree, zero-bandwidth, insufficient-neighbor, and coefficient-fit paths.
  • Match R LOESS’s even-sample bisquare MAR scale arithmetic, including extremely small residuals, while preserving the centered-MAD fallback.
  • Aligned OnlineLoess defaults across the Rust core and bindings: iterations is now 0 with the default update_mode = "incremental"; positive robustness iterations require update_mode = "full".
  • Fixed inconsistent naming of the Node.js binding as “JavaScript” across READMEs, doc-site home pages, and CITATION.cff.
  • Fixed cv_seed silently accepting negative values by validating the signed input before converting it to u64.

2.0.0

Section titled “2.0.0”

Added

Section titled “Added”
  • Added aarch64-unknown-linux-musl and armv7-unknown-linux-gnueabihf prebuilt targets with matching optional npm subpackages.
  • Added return_sorted and missing options to Loess’s SmoothOptions/StreamingSmoothOptions/OnlineSmoothOptions.

Changed

Section titled “Changed”
  • Consolidated the Node.js README (merging Installation/Documentation, dropping GitHub-only alert syntax, and removing sections covered by docs pages), renamed “When to Use” to “When to Use Batch Adapter”, and moved parameter docs into API option tables, removing parameters.md.
  • Reorganized API documentation with field tables, per-field Options sections, and a Result Structure section at the end; documented the dynamic overlap default.
  • Breaking: StreamingLoess/OnlineLoess now use dedicated StreamingSmoothOptions/OnlineSmoothOptions types instead of Batch’s SmoothOptions, dropping Batch-only fields (confidence_intervals, prediction_intervals, return_se, cv_*) from both and return_diagnostics/return_residuals/parallel from OnlineSmoothOptions. This affects TypeScript consumers; Loess’s SmoothOptions is unchanged.

Fixed

Section titled “Fixed”
  • Fixed OnlineLoess/StreamingLoess defaults to match docs: min_points changed from 3 to 2, and update_mode from "full" to "incremental".
  • Corrected docs to state that result x values follow input order after internal sorting and mapping back.
  • Fixed the “Handling Outliers” quickstart example printing nothing with only 6 points at fraction = 0.5; bumped to 0.7 so the outlier is actually downweighted.
  • Fixed the docs homepage never showing README content.
  • Fixed TypeDoc/Starlight API-reference links that returned 404s.

1.1.0

Section titled “1.1.0”

Added

Section titled “Added”
  • Added a GitHub Pages landing page at the repository root, built from README.md via pandoc and deployed by docs.yml.

Changed

Section titled “Changed”

Fixed

Section titled “Fixed”

1.0.0

Section titled “1.0.0”

Changed

Section titled “Changed”
  • Moved the tutorials/ pages into a new user-guide/use-cases/ section.
  • Split StreamingLoess/OnlineLoess content into dedicated API reference pages and standardized API examples with expected output comments.
  • Breaking: Renamed OnlineOutput’s smoothed and std_error fields to y and standard_error.

0.9.0

Section titled “0.9.0”

Added

Section titled “Added”

  • Added the Node.js binding.

  • Added weighted Streaming/Online updates, Online window diagnostics, and prediction from the current window.