Forrest Laine
@forrestlaine.bsky.social
Assistant Professor of Computer Science at Vanderbilt University. Interested in optimization, game theory, machine learning, healthcare, and autonomy, among other things.
Nice note. DP < adjoint method < solving for KKT conditions directly which is basically what any non-convex opt solver is doing anyway. write the control problem down as non-convex opt, take care with sparsity, use good LA package and 99% time it will be faster and more extensible than custom codes
November 21, 2024 at 7:28 PM
Nice note. DP < adjoint method < solving for KKT conditions directly which is basically what any non-convex opt solver is doing anyway. write the control problem down as non-convex opt, take care with sparsity, use good LA package and 99% time it will be faster and more extensible than custom codes
My setup is to lecture on my iPad hooked up to the projector. I open a zoom meeting, share screen and record to cloud. After class I upload the pdf and the zoom recording. To me feels like a classic chalkboard lecture but with the advantage of digitalization. students haven’t complained yet!
November 21, 2024 at 3:06 PM
My setup is to lecture on my iPad hooked up to the projector. I open a zoom meeting, share screen and record to cloud. After class I upload the pdf and the zoom recording. To me feels like a classic chalkboard lecture but with the advantage of digitalization. students haven’t complained yet!