You can find here code I use for my regularization experiments. All described on my blog: www.ai-experiments.com
-
r-dropout.py is dealing with dropout and spatial dropout layers. For experiments results please visit:
https://ai-experiments.com/use-your-spatial-dropout-layers-wisely/ -
r-l2.py - it a simple experiment with l2 regularization value on a very small model. Post is in progress.
https://ai-experiments.com/keras-l2-regularization-machine-learning/