solve_LSAP: Solve Linear Sum Assignment Problem

View source: R/lsap.R

solve_LSAPR Documentation

Solve Linear Sum Assignment Problem


Solve the linear sum assignment problem using the Hungarian method.


solve_LSAP(x, maximum = FALSE)



a matrix with nonnegative entries and at least as many columns as rows.


a logical indicating whether to minimize of maximize the sum of assigned costs.


If nr and nc are the numbers of rows and columns of x, solve_LSAP finds an optimal assignment of rows to columns, i.e., a one-to-one map p of the numbers from 1 to nr to the numbers from 1 to nc (a permutation of these numbers in case x is a square matrix) such that ∑_{i=1}^{nr} x[i, p[i]] is minimized or maximized.

This assignment can be found using a linear program (and package lpSolve provides a function lp.assign for doing so), but typically more efficiently and provably in polynomial time O(n^3) using primal-dual methods such as the so-called Hungarian method (see the references).


An object of class "solve_LSAP" with the optimal assignment of rows to columns.


Walter Böhm kindly provided C code implementing the Hungarian method.


C. Papadimitriou and K. Steiglitz (1982), Combinatorial Optimization: Algorithms and Complexity. Englewood Cliffs: Prentice Hall.


x <- matrix(c(5, 1, 4, 3, 5, 2, 2, 4, 4), nrow = 3)
solve_LSAP(x, maximum = TRUE)
## To get the optimal value (for now):
y <- solve_LSAP(x)
sum(x[cbind(seq_along(y), y)])

clue documentation built on Nov. 19, 2022, 5:05 p.m.