Využití neuronových sítí pro prostorové interpolace

Abstract

The bachelor thesis deals with the useing of neural networks for spatial interpolation. The thesis describes the theoretical basis of this method. Afterwards, it is used for interpolateing the surface of the selected area. The backpropagation algorithm was used for learning of Neurons. The usability of the selected software has been tested, experience in using it are mentioned. The calculation was performed for neural networks with different neurons, the range of neurons was taken from the literature. The accuracy of results for individual networks and the time necessary for the calculation were monitored. Dependence on the accuracy of the result on the number of neurons was found. To compare these results, calculations were performed on the same territory using routine methods of kriging and inverse distance weighted. The results of these methods were compared to the results of the best neural networks. Based on a comparison of the results, the conclusion about the approach of the neural network method to the normal use is presented.