Testing of NN after training.

Hello everyone

Firstly, I am presently working with ANN for a fitting problem to map between data set of numeric inputs and a set of numeric targets output. I’m using Matlab version 7.11.0.584 (R2010b) ANN tool box (nntool). I trained ANN with input data (84x294 matrixes) and target data (1x294 matrix).The algorithm itself dividing the data into training, validation and test sets. I carefully followed the steps in nntool box to create and train the network by selecting appropriate network training parameters. Finally, I obtained confusion plots of performance, trainstate and regression. My query is how to test the network after training in order to know the quality of my model with aforementioned input and target data? Again, how to extract the data for each of this plot; by this I mean training, validation and test data from the result of ANN plots. I will be glad if anyone can send a link of appropriate literatures on the confusion plots of ANN as well.

Secondly, following the same procedure as explained above, in this case I divided the data into training and test set by random selection. Total number of samples = 294 and total number of elements = 84 and target = 294. Following random selection criteria, 80% was used to build the model whereas 20% for the testing. Training (84x 236 matrixes) while testing (84x58 matrixes). Query, do I need to upload all the targets during the network training or I need to split it too and upload only the one for the training data?

Thank you as I look forward to kind and positive response.

Best wishes

[R R MAHESH]

Brother can you help me with this i dont know much about matlab, i am a civil student, i am doing data prediction using neural network, i have 133 data sets of both input and target data out of this i want to train only 106 datas but when i train in the regression curve it is showing the testing set and validation set also. so kindly if u can teach it woulld be very helpful. my email is [e-mail]. u can contact me through this if possible. It would be very helpful