Gradient descent algorithm are currently used to solve inverse problems and to optimize transducers configuration.
Two main algorithms are implemented:
Systematic Gradient Descent (SysGD)
Stochastic Gradient Descent (StoGD).
An introduction to these algorithms and a basic implementation built from scratch in Python is available in the material presented at the PyData Madrid Conference (8-10 APril 2016). Talk, code and ipython notebook are available at the GitHub repo: