how do i sub-select and make a stack of fully connected layers into partially connected layers?
1 回表示 (過去 30 日間)
古いコメントを表示
take the following example of a network architecture:
layers = [
imageInputLayer([20 20 1],'Name','input')
fullyConnectedLayer(100,'Name','fc1')
fullyConnectedLayer(50,'Name','fc2')
fullyConnectedLayer(30,'Name','fc3')
fullyConnectedLayer(10,'Name','fc4')
softmaxLayer('Name','softmax')
classificationLayer('Name','classOutput')];
I'd like to do the following
- train such a network on some custom dataset
- analyze all the weigths,
- drop out some weights (make them zero) using some criterion (not important for now) and
- design a new network with this new "connectivity pattern" - i.e. i want to completely remove the possibility of training these weights - i would like to remove these connections.p.s. these new layers would be "partially connected" and not fullyConnected.
- train this novel network architecture from scratch but this time id have fewer weights to learn.
I would like help in trying to implement this. Thanks in advance
回答 (0 件)
参考
カテゴリ
Help Center および File Exchange で Image Data Workflows についてさらに検索
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!