In this notebook I look at doing soft shrinkage as introduced by Donoho and Johnstone in 1994. I then see if it is possible to learn the soft shrinkage thresholds with backpropagation rather than using Bayes risk.

Check it out here.