Version 5.0.0

• Removed all deprecated code.
• Added external fuzzy cluster validity indices.
• Fixed GAK's sigma estimation in the non-proxy version. The estimation routine in the proxy version also changed slightly, and the estimates will be different due to randomness.
• Fixed the calculation of the Silhouette CVI (#21).
• Modified tslist to hopefully make it faster.
• Partitional and fuzzy controls (tsclust-controls) now specify version 2 of the main loop by default.
• Packages clue and ggplot2 are no longer automatically attached by default.
• Fixed minor potential bugs in several functions.
• Changed plyr dependency for dplyr.
• Minor vignette updates.

Version 4.1.1

One commit was added to version 4.1.0 to fix the last of the following changes:

• Added some fuzzy cluster validity indices to cvi.
• When using TADPole clustering through tsclust, custom parameters can now be passed to a custom centroid function if provided.
• Modified the partitional/fuzzy loops, which were doing an extra iteration. Previous version still available for backwards compatibility. This modification was wrong in version 4.1.0.

Version 4.1.0

• Added some fuzzy cluster validity indices to cvi.
• When using TADPole clustering through tsclust, custom parameters can now be passed to a custom centroid function if provided.
• Modified the partitional/fuzzy loops, which were doing an extra iteration. Previous version still available for backwards compatibility.

Version 4.0.3

• Deactivating PAM precomputation will no longer use sparse matrices by default, explicitly set pam.sparse to TRUE if you want this functionality.
• Fixed multivariate plots (#18).
• Fixed zscore for data frame input (it was still coercing column-wise instead of row-wise).
• Added an additional vignette with some timing experiments.

Version 4.0.2

• Ported the proxy versions of lb_keogh, lb_improved, SBD, GAK and dtw_basic to C++ and improved them by using the bigmemory package.
• Ported TADPole to C++.
• Ported part of the algorithm that updates sparse distance matrices to C++.
• Improved the optimizations for symmetric matrices that are calculated in parallel.
• Fixed tsclustFamily's dist function for matrix or data frame input.
• Fixed partitional PAM centroids for dtw_lb distance and pam.precompute = FALSE (#16).
• Exported a function to coerce matrices and data frames to a list.
• Updated documentation.

Version 4.0.1

• Ported dtw_lb to C++ when using dtw_basic.
• Modified some tests to account for rounding error (CRAN request).

Version 4.0.0

• Optimized TADPole for multiple k and dc values.
• Optimized PAM centroids with pam.precompute = FALSE by using sparse matrices from the Matrix package.
• Optimized shape_extraction by using the eigs_sym function from the RSpectra package.
• Implemented the DTW lower bounds in C++.
• Implemented DBA in C++.
  • Implemented an alternative version of multivariate DBA that might be faster. See its documentation.
• Added a symmetric control for fuzzy clustering.
• Partitional, hierarchical and fuzzy configurations in compare_clusterings now take into account the symmetric control parameter.
• The functionality for pick.clus in compare_clusterings changed depending on the value of return.objects.
• Fixed an error that sometimes caused objects returned by tsclust to have duplicated elements in the args slot.
• Fixed DTW symmetry detection for fuzzy clustering.
• Some internal functions changed, so older objects might no longer be compatible. Try using update(old_TSClusters_obj).
• The dtwclust function is now deprecated. Try using as(dtwclust_class_obj, "TSClusters") for old objects.
• Changed name of function compute_envelop to compute_envelope (old one still available but deprecated).
• Minor vignette updates.
• Several internal optimizations.

Version 3.2.0

• Added functions compare_clusterings and helpers to compare many clustering configurations, possibly in parallel.
• Fixed an error in tsclust that prevented CVIs to be calculated for hierarchical/TADPole cases if a custom centroid function was used.
• Added slot seed to the objects returned by tsclust.

Version 3.1.2

• The arguments in tsclust's ellipsis are now passed to all preprocessing, centroid and distance functions.
• Fixed some symmetry detection in tsclust when using DTW.
• Updated vignette to use tsclust in the examples.

Version 3.1.1

• Seeds are now set when calling TADPole through dtwclust with a parallel backend and multiple values of k, in case the centroid function has randomness associated.
• Added a new experimental function tsclust that should be functionally equivalent to dtwclust but is hopefully more coherent in general.