Closed bjack205 closed 2 years ago
Title needs work. I propose something like "JDMD: Jacobian-Regularized Dynamic Mode Decomposition for Efficient Data-Driven Control"
Overall, I think the scope is good. The paper is a little verbose, especially in the background sections, and can be tightened up a lot, but I'll wait to do that until I have some more time. A few high-level comments for now: 1) The "lifted" vs. "projected" story is not clear. I would just stick with whichever one works best and drop the other one. 2) Notation in Sec. 3 is not very clear. Instead of (4), set up the optimization problem and make it clear exactly what variables you are optimizing etc. 3) More/better visualizations in the examples section: Can we put something in to visualize the trajectories? Freeze-frames from animations? You want to have that stuff in there so that a casual reader can skim, see the graphics, read the figure captions, and know what's going on.
Overall, I think the scope is good. The paper is a little verbose, especially in the background sections, and can be tightened up a lot, but I'll wait to do that until I have some more time. A few high-level comments for now: 1) The "lifted" vs. "projected" story is not clear. I would just stick with whichever one works best and drop the other one. 2) Notation in Sec. 3 is not very clear. Instead of (4), set up the optimization problem and make it clear exactly what variables you are optimizing etc. 3) More/better visualizations in the examples section: Can we put something in to visualize the trajectories? Freeze-frames from animations? You want to have that stuff in there so that a casual reader can skim, see the graphics, read the figure captions, and know what's going on.
First draft of the paper for Zac's review.
@jhlee0997 don't merge this. I'm just opening it to allow for some open discussion.