We mainly study the application of neural network algorithm on permanent estimation problem.We analyze the advantages and disadvantages of convolution neural network and artificial neural network in the estimation problem.The influence of different network structure,different training data sets and other factors on the prediction accuracy and the time complexity is well discussed.Through experiments,we show that for the 8 order to 18 order matrix,the HF-CNN structure of the ANN based on characteristics can make effective estimation on the logarithm of permanent in the case of low time complexity.We further improve the estimation accuracy of permanent by combining other algorithms. |