Skip to contents

Changelog

spacefinder 0.2.2

Bug Fixes

  • Fixed issues in coef() method for Box Learner when only one hyperparameter is present
  • Fixed issue in coef() method for Box Learner with setting vectorize = TRUE and degenerate case, where min=max.
  • Fixed faulty normalization of weights within augment().
  • Altered synthetic dataset benchmark_data.

spacefinder 0.2.1

Internal changes

spacefinder 0.2.0

Major Changes

  • Comprehensive roxygen2 documentation for all exported functions and classes
  • Added four vignettes: getting-started, learner-comparison, density-estimation, categorical-hyperparameters
  • Added benchmark_data dataset for examples and vignettes
  • Standardized transformation convention: y = Ax + b maps from fitted subspace to unit hypercube

New Features

  • augment() methods for Box and Polygon learners to fit Beta densities
  • coef.LearnerSubspaceBox(vectorize = TRUE) now returns transformation matrices A, b

Breaking Changes

  • LearnerSubspaceElips is now called LearnerSubspaceEllipsoid
  • LearnerSubspaceBoxGeneral is now called LearnerSubspacePolygon

Bug Fixes

  • Fixed data.table NSE issues in coef() methods
  • Fixed formula parsing in TaskSubspace initialization
  • Corrected transformation in augment.LearnerSubspacePolygon()
  • Fixed roxygen2 cross-reference warnings

Documentation

  • Complete documentation for all R6 classes (LearnerSubspace, TaskSubspace, and subclasses)
  • Documented all S3 methods (coef, augment, autoplot, summary)
  • Added mathematical notation and detailed algorithm descriptions
  • Cross-referenced related functions throughout

spacefinder 0.1.0

  • Initial release
  • Implemented LearnerSubspaceBox, LearnerSubspaceElips, LearnerSubspaceOrientedBox
  • Added methods: coef(), summary(), autoplot(), outliers()

Features