Selection of the most "efficient" shortened Reed-Solomon code from a neural network database

The catalog of Reed-Solomon (RS) codes is a rather long one. To select a proper code for a given application, the system designer is compelled to deal with numerous tables, graphs and equations. We have reported our result of designing an artificial neural network (NN) from which one can select the most "efficient" unmodified RS code for a specific application. In this article we present the continuation of our work, in development of an artificial NN database for selection of shortened RS codes for a given application. A student version of the MATLAB Neural Networks Toolbox is used for NN simulation. The Levenberg-Marquardt learning algorithm is used to train the NN. The resultant NN has five inputs, nine units in the hidden layer, and two units in the output layer. The outputs are the shortened "n" and "k". The test data results show the accuracy of selecting the correct code length and code dimension is 84.4% for shortened codes.