ap: assignment problem solvers for go

This package provides interfaces and data structures common to formulating and solving assignment problems, as well as codes for solving particular variants. More details about these can be found in:

Rainer Burkard, Mauro Dell'Amico, and Silvano Martello.
"Assignment Problems - Revised Reprint."
Society for Industrial and Applied Mathematics (2012).

At this time, ap only provides an incremental code to solve the Linear Sum Assignment Problem. Additional forms are planned for future milestones.

LSAPs take the following form:

min   ∑_i,j c_ij * x_ij
s.t.  ∑_i   x_ij = 1      ∀ j
      ∑_j   x_ij = 1      ∀ i
            x_ij ∈ {0,1}  ∀ i,j

Quick Start: CLI

To solve LSAPs from the command line, first install the lsap binary using Go.

go install github.com/ryanjoneil/ap/cmd/lsap

lsap reads JSON input data in the form of a square cost matrix from standard input and writes an optimal permutation and cost to standard output.

cat <<EOF | lsap | jq
[
    [  90,  76,  75,  70 ],
    [  35,  85,  55,  65 ],
    [ 125,  95,  90, 105 ],
    [  45, 110,  95, 115 ]
]
EOF

{
  "assignment": [
    3,
    2,
    1,
    0
  ],
  "cost": 265
}

Quick Start: Packages

Examples are available in the package docs.

ap: assignment representations & interfaces

Package ap provides solution representations and interfaces for working with assignment problems and solvers.

go get github.com/ryanjoneil/ap

The default representation of an assignment produced by an Assigner is a Permutation.

a := SomeAssigner{} // implements ap.Assign
p := a.Assign()     // p is an ap.Permutation

Permutations can be converted to cyclic and matrix representations of assignments, and vice versa. All representations provide Inverse methods reverse the direction of assignment.

p := ap.Permutation{1, 0, 2, 6, 5, 3, 4}
p.Cycles()  // {{0, 1}, {2}, {3, 6, 4, 5}}
p.Inverse() // {1, 0, 2, 5, 6, 4, 3}
p.Matrix()  // p[u] == v -> m[u][v] == true

ap/lsap: linear sum assignment problem solver

Package ap/lsap provides a efficient, iterative implementation of a primal-dual linear sum assignment problem solver.

go get github.com/ryanjoneil/ap/lsap

LSAPs are easy to construct and solve from a cost matrix.

a := lsap.New([][]int64{
    {10, 15, 12},
    {51, 75, 23},
    {11, 91, 10},
})

permutation := a.Assign() // [1 2 0]
cost := a.Cost()          // 49

lsap provides command line flags for outputting dual prices and reduced costs.

lsap -h

lsap solves linear sum assignment problems, given a square cost matrix
Usage:
        lsap < input.json -dual -rc > output.json
        cat <<EOF | lsap | jq
        [
                [  90,  76,  75,  70 ],
                [  35,  85,  55,  65 ],
                [ 125,  95,  90, 105 ],
                [  45, 110,  95, 115 ]
        ]
        EOF
Flags:
        -cycles
                output cyclic assignment form
        -dual
                output dual prices
        -matrix
                output matrix assignment form
        -rc
                output reduced cost matrix