A Learning Method based on Hopfield Neural Network and Its application in Point-feature Labeling Placement Problem

This paper proposes a new learning method based on Hopfield Neural (HN) network to optimize the Point-Feature Labeling Placement (PFLP) problem. The learning method attains a balance between penalty function and original objective function based on the principle of a physical weight balance, and can converge to a solution with better stability. This improved algorithm also allows HN network to be competitive among other traditional algorithms such as genetic algorithm and simulated annealing algorithm when solving the PFLP problem and other constrained problems..