Skip to content

NEWS.md

Latest commit

 

History

History
102 lines (80 loc) · 3.77 KB

NEWS.md

File metadata and controls

102 lines (80 loc) · 3.77 KB
Raw

fdacluster (development version)

fdacluster 0.4.2

  • Updated C++ code to use std::isfinite() on scalar values instead of arma::is_finite() which is meant for Armadillo objects. Also, replaced use of arma::is_finite() in favour of .is_finite() method of Armadillo objects.
  • Better handling of precomputed data in vignettes.
  • Skip examples and vignettes to accommodate some weird issues on some Rhub platforms.
  • Fixes #17 which reported integer overflow when computing linear indices for large datasets by computing on the fly indices (i,j) from linear indices.

fdacluster 0.4.1

Bug fixes

  • Properly set up future workers by ensuring that fdacluster is loaded.
  • Replace SRSF acronym with the correct SRVF one.

fdacluster 0.4.0

Major features

  • Expanded arguments of fdakmeans() to allow for more control over the type of input functional data:
    • is_domain_interval allows one to state if all curves are defined on the same fixed interval;
    • transformation specifies the transformation to be applied to the data before clustering.
    • check_option_compatibility() handles errors when incompatible options are selected.
  • Created two separate C++ classes for $L^2$ distance and normalized $L^2$ distance; the former cannot be used in combination with dilation or affine warping classes because it is not invariant to these transformations.

Minor improvements and bug fixes

  • Integrated distances in C++ classes are now computed via arma::trapz().
  • Added talk given at Rencontres R 2023 in Avignon, France to the News section of the website.
  • Reduced number of dependencies: removed dplyr, forcats, tidyr, purrr.
  • Replaced furrr dependency in favor of future.apply to further reduce number of dependencies.
  • Updated README file.
  • Updated GHA workflows.
  • Updated vignettes.
  • Bug fixes.

fdacluster 0.3.0

  • Added median centroid type;
  • Median and mean centroid types are now defined on the union of individual grids;
  • Simplified caps class to avoid storing objects multiple times under different names;
  • Added vignette on initialization strategies for k-means;
  • Added article on use case about the Berkeley growth study;
  • Added article on supported input formats.

fdacluster 0.2.2

  • Make sure one can use fdacluster with namespace notation.
  • Make sure not to use fda or funData before checking it is available.

fdacluster 0.2.1

  • Add DBSCAN clustering;
  • Fix C++ compiler issues that errored when accessing empty vectors.

fdacluster 0.2.0

  • Add hierarchical clustering;
  • Enforce n_clusters in output via linear programming (LP) using the lpSolve package;
  • New caps class for storing results from functional Clustering with Amplitude and Phase Separation in a consistent way;
  • Add tools for comparing clustering results (mcaps objects, autoplot and plot specialized method implementations);
  • Add seeding strategies for kmeans (via hierarchical clustering or k-means++ or k-means++ with exhaustive search of the first center or exhaustive search of all the centers);
  • Add within-cluster domain auto-extension via mean imputation;
  • Add possibility to cluster according to phase variability instead of amplitude variability.
  • Renaming of functions: to perform k-means with alignment, now use fdakmeans(), to perform HAC with alignment, now use fdahclust().

fdacluster 0.1.1

  • Fixed undefined behavior sanitizer issues spotted by UBSAN.
  • Added reference to published work related to the package in DESCRIPTION.

fdacluster 0.1.0

  • Initial release.
  • Added a NEWS.md file to track changes to the package.