why my neural network is not training properly when i am using traingdm or traingd as a training function?

sir, i am using 'traingdm' (Gradient descent with momentum backpropagation) in my neural network as a training function but the mse(mean square error) is not decreasing during training.when i am using 'trainlm' or 'trainbr' as a training function its giving me pretty good result but i am interested to use 'traingdm' as a training function.i cant understand whats wrong with my network and code.i am inserting my performance plot and code.

    % code
  net = feedforwardnet(20,'traingdm');
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;
[net,tr] = train(net,x,t);

[Greg Heath]

You are going to have to provide the input and target