Mathematics of Deep Learning
Spring 2020

End-to-end training of deep visuomotor policies (Guided Policy Search)


This is what it’s all about! Guided policy search is one of the most efficient techniques for robotic control from vision with partially known environments. This week we’ll put it all together, showing how GPS combines trajectory optimization, imitation learning, and constrained optimization to find high-quality neural network policies with very little real-world experience.


Required reading

Optional reading