On the Levenberg-Marquardt training method for feed-forward neural networks

摘要

The Levenberg-Marquardt (LM) training method is the most effective method for feed-forward neural networks with respect to the training precision. This method is well-known and popularly described in the neural networks literature. Nevertheless its implementation contains some difficulties because of the specific shape of the cost function and the large amount of variables. Here we give in sufficient details an example of a program implementation of the LM training method. This implementation (as a Matlab application) seems to work well with various examples.