how to train a neural network with a matrix as input

For now, I have got my feature matrix( x_train), which is 25 * 1000, each column is one train data and each train data has a dimension of 25 feature.

Also, I get my target matrix(y_train), which is 27 * 1000, each column is the corresponding label of the train data. Each column looks like [1 0....0] or [0 1 0...0] or [0 0 1 ... 0], because there are 27 different kinds of labels, and I try to represent them as vector.

Then, I create a feedforward neutral network with

net = feedforwardnet(10)

and train by

net = train(net,x_train,y_train)

After that, I try to make a prediction with this trained net. However, it gives a result like

    0.0407
    0.2103
   -0.7430
    2.1438
    0.2586
    0.1181
   -0.2998
    1.0055
    0.3115
   -0.7575
   -0.7809
    0.7042
    1.0576
   -0.4813
    1.2649
    1.0238
    0.0825
    0.1803
    1.3886
    0.5525
   -0.4226
    1.0090
    1.8860
   -0.1083
   -0.2245
    1.2130
    0.2512

In my expection, it should output something like the labels, [1 0...0] or [0 0 0 0 1 0....0].

How can I do it correctly?