Hacker News Reader

Comment by RhysU | Hacker News Reader

> We report on a data-driven method for learning a nonperturbative guiding center model from full-orbit particle simulation data.

> Then we describe a data-driven method for learning from a dataset of full-orbit α-particle trajectories. We apply this method to the α-particle dynamics shown in Fig. 1 and find the learned non-perturbative guiding center model significantly outperforms the standard guiding center expansion. Our proposed method for learning applies on a per-magnetic field basis; changing requires re-training.

Is this interpolation at its heart? A variable transformation then a data fit?

Anyone know which functionals of these orbits are important? I don't know the space. I am wondering why the orbits with such nuance should be materially important when accessed via lower-order models.

View on HN

Haven’t read the article yet, yet alone the paper but based on what you’ve quoted these are ongoing challenges with regards to confinement. Think tokamak vs stellarator. Magnetoplasmahydrodynamics is hard because you have all the complexities of the navier-stokes combined with Maxwell and thats just scratching the surface. Sensitive dependence on initial conditions has never been so sinister as in plasma confinement. Orbital perturbations quickly lead to turbulent instabilities which lead to containment breach which can lead to multi-million degree hyper velocity jets tearing a hole through your multi-billion dollar toy in seconds.

Are these sorts of instabilities harder to control in a tokamak as compared to a stellarator, or did you just bring those up as examples of magnetic confinement?

I’ve since gone back and read the article, haven’t looked for the paper yet.

> Are these sorts of instabilities harder to control in a tokamak as compared to a stellarator, or did you just bring those up as examples of magnetic confinement?

I was just shooting from the hip in the earlier comment alluding to your question and bringing them up as examples of two different approaches to addressing the said issues with instabilities that lead to the complexities of confinement. I think its just a terribly fun thing to think about because of its complexity. Stellarators are attempting to solve the issues passively through design. Tokamaks on the other hand with active control. Theres trade offs to both and neither has reached break even output yet.

I’m personally largely bored with them and think linear is the way to go, even though the laser based inertial confinement reactor at Lawrence is the first to reach breakeven output… experimentally at least.

Reply on news.ycombinator.com