NEWS

OSCARS 0.2.1 (2026-09-22)

oscars() gained a progress argument (default TRUE) that draws a console progress bar over the nfmax function evaluation budget, showing the percentage of that budget used and the elapsed run time. poscars() workers run with progress = FALSE.
New import of the cli package.

Added poscars(), a parallel wrapper around oscars() to run several independent OSCARS-II searches at once (one per processor core) and returns the best result. The number of cores is controlled by the new ncores argument. By default the evaluation budget nfmax is divided among the workers, cutting wall-clock time by roughly a factor of ncores; set divide.budget = FALSE for a best-of-ncores multi-start instead.
poscars() gained a seed argument for reproducible parallel runs and a cl argument for reusing a user-supplied cluster across calls.
New import of the base parallel package.
New code oscarsQN() which runs a version of oscars with intermittent quasi-Newton steps for improved accuracy and improved performance in high dimensions when there are relatively few local minimizers.

Initial CRAN submission. Initial submission contains a single optimization routine, oscars.