I'm experimenting with open research notes now I'm a free agent. I thought this might be a fun project to share, because it ended up being a small metaphor for every ML project I've ever done. Not a serious software project, just me thinking out loud.
It is fun if you ever find yourself in this situation because you can play the uno reverse card on the interviewer and ask to clarify with impenetrable jargon and look for rising panic (can I assume the graph contains a Hamiltonian circuit? etc, etc)
The error budget in the pseudorange to the satellites has various factors due to relativity, but they are just lumped into the errors in the least squares problem that you solve to get the position estimate (as per the article).
So relativity is important, but you don't need to know much about it to solve for your position. Various flavours of long baseline/network RTK will need more sophisticated modelling though.
I'd say a fairly large percentage would be disappointed that we let a citizen get treated like that and we did nothing as a country to assist, independent of anything else. Maybe I am out of touch though.
Here in Australia the coldest state (tas) uses more energy (8600) than anywhere else (e.g. qld 5500). They have basically 100% hydro power and use reverse cycle AC for heating. Not sure what else is going on there.
I'd say it's not new. Take fluid dynamics as an example, the navier stokes equations predict the motion of fluids very well but you need to approximately solve them on a computer in order to get useful predictions for most setups. I guess the difference is the equation is compact and the derivation from continuum mechanics is easy enough to follow. People still rely on heuristics to answer "how does a wing produce lift?". These heuristic models are completely useless at "how much lift will this particular wing produce under these conditions?". Seems like the same kind of situation.
Maybe progress forward will look like producing compact models or tooling to reason about why a particular thing happened.
It seems like a good application of the technology to a real problem, it just sucks so much that the meta problem is so diabolical. Definitely a local optimisation far from the global.
The whole process comes from a time when publishing was expensive, should be cheap as chips now. The system needs a rethink so "quality" can somehow bubble up to the surface given mass publication is so simple.
The biggest one I have run into is that the bandwith doesn't need to be contiguous. If you know the the important bits of the signal are contained in certain frequency bands then you can get much lower sampling rates. Basically why things like l1 reconstruction work.
It's not that long ago in human history that basically none of the jobs we do now existed. So it is kind of myopic to think that any current career is a calling. Art can become a craft again, not a career. There is nothing wrong with that.
Just for a bit of background, transmission companies want to know the temperature of transmission lines because the dynamic clearance of the line to the ground as the current heats the wire limits the max current. Surprisingly complex to model accurately as it is quite dynamic (sometimes the conductor is pinned with the insulator and it swings, or sometimes it can slip through, etc, etc).
I was also incredibly optimistic about this approach in the past but life experience has shifted me away from it. I personally can't understand people not wanting to know. It is weird dynamic. I changed jobs a few years ago and went from management to an IC role for a while. It was nice to focus on technical problems but it also did my head in. I felt incredibly neurotic knowing that stuff was going on but I was in my little bubble of productive calm.
My understanding (that might be out of date) is that the tools are weak. Ideally you would have tabular data and it would give you a digraph for the causal structure between variables. You can try this but the tools don't work reliably yet. Otherwise everyone would use them.
It is a weighed average of the propagated uncertainty of a state and the uncertainty of a measurement of that state. If you walk towards a wall with your eyes closed as time on you will be less and less certain where you are (prediction steps) then when your fingers touch the wall you will suddenly be much more certain where you are (measurement update). The linear algebra is just how you calculate the weights for a linear dynamical system with gaussian noise.