added clipper::Params::rounding param --> can be NONZERO, DSD, or DSD_HEU (default)
NONZERO: any nonzero element of u (which is in positive orthant) is taken to mean that node should be selected
DSD: select elements based on NONZERO, and then run exact dsd solver
DSD_HEU: select elements based on NONZERO, and then run heuristic dense subgraph solver as described in ICRA'21 CLIPPER paper. Note that this heuristic approach (select the k=round(max eig) largest nodes) tends to reduce recall in favor for precision.
expose clipper::dsd::solve via python, note that it can take an optional parameter of nodes --> causes dsd to search on restricted subgraph as defined by those nodes
Implements dense edge-weighted subgraph solver based on Goldberg's maxflow-mincut based algorithm
clipper::Params::roundingparam --> can beNONZERO,DSD, orDSD_HEU(default)NONZERO: any nonzero element of u (which is in positive orthant) is taken to mean that node should be selectedDSD: select elements based onNONZERO, and then run exact dsd solverDSD_HEU: select elements based onNONZERO, and then run heuristic dense subgraph solver as described in ICRA'21 CLIPPER paper. Note that this heuristic approach (select thek=round(max eig)largest nodes) tends to reduce recall in favor for precision.clipper::dsd::solvevia python, note that it can take an optional parameter of nodes --> causes dsd to search on restricted subgraph as defined by those nodes