matlab - Concept of validate for neural network -

i have problem concept of validation nn. suppose have 100 set of input variables (for example 8 input, x1,...,x8) , want predict 1 target(y). have 2 ways use nn: 1- use 70 set of data training nn , use trained nn predict other 30 sets of target validation , plot output vs target 30 sets validation plot. 2- use 100 sets of data training nn , divide outputs 2 part (70% , 30%). plot 70% of outputs vs corresponding targets training plot. plot other 30% outputs vs corresponding targets validation plot

which 1 correct??

also, difference between checking nn new data set , validation data set??

thanks

you cannot use data validation, if has been used training, because trained nn "know" validation examples. result of such validation biased. sure use first way.