A novel neural network for solving multicriteria shortest path problem

This paper studies the multi-criteria shortest path problem under resource constraints. By designing a wave-based neural network (WBNN), the global optimal solution of the problem which is difficult to obtain by traditional methods is obtained. In the design of WBNN, the overall network topology is regarded as a neural network, and each node is regarded as a wave-based neuron. The proposed wave-based neuron consists of five parts: input, wave decoder, wave generator, wave encoder and output. The first two parts are used to decode the wave from the previous neuron, and the last three parts are used to generate waves for subsequent neurons. Experimental results show that compared with other methods, the proposed method not only has reasonable calculation time, but also can obtain high-quality solutions.