Image feature

[fatemehkalantari1993]

I use 40GB RAM to run your code, and I also run the training part for each layer separately, but since the size of the image features is very high (especially in the initial layers), I am facing a memory error. What was the RAM of your system to run this code (featdec_deeprecon_500voxel_vgg19_allunits_fastl2lir_alpha100_train)???
If I want to reduce the number of units in each layer, how should I do it? I don't understand how you can select a single group per layer. I used a pre-trained VGG (imagenet-vgg-verydeep-19.mat) and MatConvNet for image features.