Notes

Some things I think are cool! Currently mainly for my own reference, so heads up and apologies for poor writing, but hopefully it will get better. Recently this looks like deep learning, RL, optimization, and robotics.

KL Estimators in RL

Different KL estimators, buggy gradients, and when to use each

#machine learning

Dual projection

Polytope projection, Lagrangian dual, normal cones, and a small interactive μ-slider.

#optimization

Draft topics

  • TBD

Thanks for stopping by :)