NEWS

HyRiM 28.08.2017

FEATURE: plot.mosg now takes arguments xlim and ylim to make all plots comparable by identical axes ranges
BUGFIX: improvement to the implementation of density (now returns zero only for categories as they are specified; nothing more)
BUGFIX: print.mosg did not list the goals by name (now does)

BUGFIX: games with numeric payoffs fail if one of the payoffs is an integer power of 10 (so that the internal distributions become degenerate in this case)

BUGFIX: density, when applied to continuous distributions, returned zero for vector arguments in some occasions

BUGFIX: preference function returned wrong results for relations between numbers and distributions

FEATURE: added vignette paper (displayable via 'RShowDoc("vignette", type="pdf", package="HyRiM")')

BUGFIX: preference among discrete distributions with different ranges but equal support was occasionally computed incorrectly

BUGFIX: error in "moment" function; thankfully submitted (with correction) by Ali Alshawish

BUGFIX: "range" now properly updated after smoothing a density (previously, the range was set based on the data, but not updated after smoothing, which caused errors when moments were computed (visible by warning messages))
FEATURE: if loss distributions with empty categories are supplied, the error now tells where the problem occurred in the game (row, column and goal)

BUGFIX: added tolerance parameter "tol" to function mgss, to account for roundoff errors behind the 5th digit after the comma when solving the linear programs

CHANGE: fixed dependency on "orthopolynom", due to some checking issues with this package (published on https://cran.r-project.org/web/checks/check_results_orthopolynom.html).

CHANGE: replaced fictitious play to compute approximate security strategies by an exact algorithm using linear programming. This will be version 2.0.0 of the package, pending for release over CRAN. It is no longer downwards compatible regarding the function codemgss</code> to code using version 1.x.x. Thanks to Ali Alshawish and Vincent Bürgin for reporting the convergence issues with fictitious play that led to this change.
FEATURE: added support for real-valued multi-criteria games that use lexicographically ordered payoffs
FEATURE: extended the function "disappointmentRate" to take matrix games over the reals and compute the disappointment rate directly for them, or for a given equilibrium therein
FEATURE: added convenience function for extraction and replacement of parts of a game's payoff structure (in the usual R-syntax x[i,j,..] <- replacement, or x[i,j,k] for extraction)
FEATURE: added an option "cleanUp = FALSE" to plot.mosg to retain the SVG file for subsequent use elsewhere (e.g., in papers)

CHANGE: updated manual pages by adding new (recent) literature references and revising explanations

BUGFIX: in some cases, numeric roundoff errors encountered internally in the constraints of the LPs made the linear programs infeasible. Added an optional tolerance parameter "tol", defaulting to zero for exact solutions, but usable to fix the roundoff errors if needed.

BUGFIX: function "mosg" incorrectly handled the parameter setting "byrow = FALSE"
BUGFIX: "moment" was on categorical distributions incorrectly using "range" rather than "supp"