relation_dissimilarity(), relation_consensus()and relation_choice() gain methods for generalized paired comparison dissimilarities (which include symdiff, CKS, and Emond-Mason dissimilarities as special cases).
relation_is() meta-predicate function added, to replace the current single predicate functions in future releases.
is.na method for relations added.
Changes in version 0.6-6
No user-level changes.
Changes in version 0.6-5
Safeguard code against nchar(NA) being NA.
Changes in version 0.6-4
Rd improvements.
Changes in version 0.6-3
Improve namespace and avoid search path semantics.
Improve message strings.
Changes in version 0.6-2
Package metadata improvements.
Changes in version 0.6-1
Rd improvements.
Changes in version 0.6
Code added for computing (weakly/strongly) connected components, condensations and component representations.
transitive_reduction() and the plot method for relations now also work for cyclic relations.
The plot method for relations now also allows the drawing of “raw” graphs without any transformation.
Changes in version 0.5-11
Bug in transitive_reduction() fixed.
Changes in version 0.5-10
Vignette fixes.
Package news converted to Rd format.
Changes in version 0.5-9
Add predicates for (a)cyclic relations.
Fix incorrect transitive_reduction() code.
Changes in version 0.5-8
Ensure that examples do not fail when package Rgraphvizis unavailable.
Changes in version 0.5-7
support for finding up to n consensus relations added for most methods.
Changes in version 0.5-5
mean() method added for relation ensembles.
Jaccard dissimilarity added.
relation_precursors() added.
support for missings added.
relation_choice() gains a method for computing Euclidean choices and Schulze Winners.
relation_impute() added with methods for linear, weak, and partial orders.
Sparse lightweight array and matrix code moved to packageslam.
Changes in version 0.5-1
Rd fixes.
Changes in version 0.5
New relation_consensus() method for Cook-Kress-Seiford (CKS) dissimilarity.
New symdiff fitters for consensus preorders, and family ‘G’ for general crisp endorelations.
Performance enhancements for finding all symdiff consensus solutions, and symdiff consensus for equivalences or weak orders with a given number of classes.
relation_choice() gains a method for CKS dissimilarity, and aweights argument.
Performance enhancements for symdiff choice via employing a more compact linearization of the underlying binary quadratic program (BQP).
GLPK is now used as the default MILP solver.
Consensus and choice fitters now use sparse constraint matrices by default.
New class for representing rankings.
Improvements for plotting relations and relation ensembles.
New function relation_cover() for computing the covering relation of an endorelation.
New function relation_trace() for computing the left or right trace of an endorelation.
New simple choice function relation_elements() for determining the minimal, first, last or maximal elements of an endorelation.
New predicate relation_is_quasitransitive().
Changes in version 0.4
symdiff fitters can now use the GLPK, SYMPHONY and CPLEX solvers in addition to lp_solve, and optionally use sparse constraint matrices
more methods added to relation_violations(). optionally, return all tuples violating a property.
relation_scores() now also works for relation ensembles
Transitive/reflexive closure and reduction functions are now available as methods for closure and reduction generics (provided by package sets)
During creation of relations, domains are converted to sets and incidences are canonicalized to match the internal order of the domain
relation_consensus() can now compute all solutions for the Condorcet method, and gains a ‘majority’ method
subassignment method for ensembles added
set_intersection() accepts any number of arguments
as.relation() generic now has a ... argument
the attrs argument in plot.relation_ensemble() is now a list, so that the Rgraphviz parameters can be changed in each plot individually
all.equal() methods added for relations and ensembles
Changes in version 0.3
rank method is default for relation_scores()
the relations in the SVM Benchmark data are now all reflexive
fitters can now use a sparse constraint matrix representation
new symdiff fitters for the S ("symmetric") and M ("matches") families
relations are now subsettable
fitters for Cook-Seiford method and Euclidean consensus added
new summary() method for relations which computes all implemented predicates
relation_choice() for choosing “winner” objects based on an ensemble of relations between these
Changes in version 0.2
Copeland's method added to relation_consensus()
relation_classes() added to extract and pretty-print (ordered) classes from preferences and equivalences.
relation_violations() added to compute a measure of remoteness from a specified property (e.g., symmetry, transitivity, etc.)
set_symdiff() fixed for more than two elements (used to compute the complement of the union and the intersection of a set, which is wrong in general). Now, the argument list is treated recursively.
Vignette added
New symdiff fitter added for the C ("complete") and A ("antisymmetric") families of relations
Summary methods for tuples and sets added
Ops and t() methods for the relation_ensemble class added
plot.relation() fixed (used to plot the dual, now changes the graphviz layout)
Ops.tuple() now correctly handles unary operators, and preserves class
transform_incidences() added for transforming incidence matrices of endorelations from/to other codings found in literature