The sae. Ae {I}=nntrain (sae. Ae {I}, x, x, opts);
Started training network is good after the first since the encoding,
Call this function for training, I have some can't understand what this train_y specific refers to, its data is the load mnist_uint8, input data train_x is these handwritten Numbers, then train_y here what specific refers to,
Function [nn, L]=nntrain (nn, train_x train_y, opts, val_x, val_y)

Another problem is sometimes used in the training network is supervised, sometimes it is unsupervised,
1. If have supervision is labeled, so what specifically mean the tag,
2. If unsupervised training just have the training data, no label, that how to optimize parameters of
3. Adjustable parameters is based on the bp back propagation algorithm, how to calculate the residual target data y get (it) took a image
4. There is the label and the target data of calculating residual y have a contact?


Just at the beginning of the contact network of each link can not understand, hope to have a tall person to give directions