Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Computation of cross-covariance of state and action #35

Open
dvtailor opened this issue Oct 7, 2019 · 0 comments
Open

Computation of cross-covariance of state and action #35

dvtailor opened this issue Oct 7, 2019 · 0 comments

Comments

@dvtailor
Copy link

dvtailor commented Oct 7, 2019

From only looking at the docstrings of the relevant functions, I think I noticed a discrepancy to the paper. I am writing this without checking the math in the code so I may be wrong.

V returned in RbfController.compute_action() in controllers.py
corresponds to Cov[x,u]

From backtracking to MGPR.predict_given_factorizations() in models/mgpr.py, I think the docstrings indicate that:

V = cov[x,x]^{-1} @ cov[x,pi] @ cov[pi,u]

where I call pi the action before squashing

From section 5.5 of the 2015 paper, it says:

V = cov[x,pi] @ cov[pi,pi]^{-1} @ cov[pi,u]

Are these expressions equivalent or have I misread something. Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant