Almost the same but Different network structure for each sample

[Wei Xiong]

Hi all,

   I am very new to pylearn and to deep learning. I have a problem, which I would like to solve using deep learning.

I have about N=300 training sample (X_i, y_i). y_i is a real number. But X_i is comprised of k_i images. k_i is

between 100 and 200 and it is different from sample to sample. 

  I would like to build a network (typical with deep learning layers, conv, max pooling, dropout), which generate

an intermediate output o for each of the k_i image. And the final output is a summation of all the k_i intermediate 

outputs. I would like the final output to be fit to y_i.

  Using all the N=300 samples, I hope to find the weights for the layers in the network.

  Is this possible?

Thanks,

Mengda