If you have a problem that has already been partially solved and is to be solved further to obtain a better, optimal solution, information describing the solution now available may be used as an initial solution. This is called warm starting the optimization, and the supplied solution data are called the warm start.

Some data can be changed between the time when a warm start is created and when it is used as a warm start for a subsequent PROC NETFLOW run. Elements in the arc variable cost vector, the nonarc variable objective function coefficient vector, and sometimes capacities, upper value bounds, and side constraint data can be changed between PROC NETFLOW calls. See the section Warm Starts. Also, see Example 6.4 through Example 6.8 (the TV problem) for a series of examples that show the use of warm starts.