Marginal allocation algorithm for nonseparable functions

Marginal allocation algorithm is implemented to discrete allocation problems with nonseparable objective functions subject to a single linear constraint. A Lagrangian analysis shows that the algorithm generates a sequence of undominated allocations under the condition of discretely convex objective functions and Lagrangian functions. The case of separable functions is proven to be a special case. An application is provided to illustrate the method and various size randomly chosen problems are run to demonstrate the efficiency of the marginal allocation algorithm.