Neural Networks for Combinatorial Optimization

摘要

In most applications of neural network models to combinatorial optimizationproblems reproduction is used. In this paper we concentrate on reproduction approaches that apply to combinatorial optimization problems in general. In that context we deal with Boltzmann machines, Hopfield networks, and multi-layered perceptrons, In Section 2, we introduce our formulation of combinatorial optimization problems. Selection 3 discusses Boltzmann machines and Hopfield networks and their relation with combinatorial optimization. The emphasis in this section is on Boltzmann machines. In Section 4 we show how combinatorial optimization problems can be reformulated as classification problems and how they can be solved by multi-layered perceptrons. Section 5 presents a discussion and some concluding remarks.