My training data is too big for regression

The training data input is (12925*7) and my training data output is (12925*1). It takes alot of time to train the network.Similarly the testing data is (4000*7)and (4000*1). I want to do estimation(regression) using neural network and rbfnn but the matlab rbfnn function doesnot work with this dataset. its too big for it to do the required task.All the data is numerical. I am new to data mining.What should i do? Need help

Answer 1

[Greg Heath]

Randomly divide the dataset into m subsets that are small enough to train.

Run all of the data thru each of the m nets.

Rank the performances of the m nets.

If the performance of the best net is unsatisfactory, average the outputs of the best 2 nets, etc.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Greg Heath]

Since weight initialization and datadivision are random, it may take multiple runs to obtain good deigns for each of the m nets.