I am trying to train an autoencoder network. 1) I cannot upload all the training data into memory. can you show show how to successively train the same network with new data? For example, this is creation with initial data: autoenc = trainAutoencoder(trainData,hiddenSize); Now, how can I train autoenc with more data?
2) Each data point for me is 20000x4. I assume I will have 20000X4x1 weights to each of my 25 hidden nodes. How can I access those one at a time?