Why a patternet has both validation set and holdout validation set?

Hello,

I haved seen some codes about digit recognition and classification problem. They have both validation set and also hold out validation set. How are they different and why both have been used in a network?

c = cvpartition(n,'Holdout',n/3);

and

net.divideParam.trainRatio = 80/100;

net.divideParam.valRatio = 10/100;

net.divideParam.testRatio = 10/100;

Thank you so much.