What is best BP training function to train data with following input in MLP?

I have a lorenz time series data with 3*19999 input and same size of target data. I have to train this network with MLP by using delta learning rules with Back propagation. The hidden layer will use sigmoid activation function and linear function for output layer.

I want to know that which training function should I use for the network training.

Answer 1

[Greg Heath]

Defaults are

 trainlm for regression

 trainscg for classification and unit column vector targets

 trainrp for huge datasets