how to deal with out of memory error when training a large data set or large neural network narx

how to deal with out of memory error when training a large data set or large neural network stacked autoencoder ......just initiated to analyze deep neural network with time series predictions , what i guess is by using batch trainings but can any one help me in detail

Answer 1

[Greg Heath]

What are

 1. input and target
 2. size(input), size(target)
 3. Significant lags of
    a. target autocorrelation function
    b. input/target cross correlation function

Answer 2

[SAM]

i am xtremly sorry i wrote narx but i actually tried it with stacked autoencoder . hidden layers [100 150] even on HPC server i was getting same out of memory error.

inputs are 4 *2795 and output 1*2795 .