Data generalization

[himora...@gmail.com]

Hi everyone,

I want to evaluate the accuracy a CNN network with a dataset in which X1 is number of training data and X2 is number of testing data.

I would need a validation data so that the result is generalized. Therefore, i have divided the X1 to X1_train (80% of X1) and X1_val (20% of X1)

I have run different CNN architecture, training on X1_train and testing on X1_val and obtain an optimal architecture. Then, i use this architecture to train on X1 and test on X2. Is this way generalized enough?

Hope to receive your advice on my situation.

Best regards