These are blogs about trying to get neural networks to forget stuff by removing parameters only.
Introduction to Forgetting
Have to start somewhere. This is really just a short intro to these blogs.
First Experiment
It works, read about it here. I explain the method I use, and how well it works on this simple model.
First Experiment Discussion
More in depth discussion about that first experiment. What I am going to try next and why. Also talks about what this could be used for.
Targeted Outputs
In this blog I look at the output values of the model to get a better idea of what is going on when parameters are removed. This was more interesting than I thought it would be, also longer.
Next
Next I will be looking at different classes. There is a difference in how different classes can be forgotten. This might give me some interesting insight into how information is stored in the model.
Later
Later I will write a full summary of all of these, but I have not finished yet.