3 ms·
Largely superseded? This is a sarcastic comment, right? In case it isn't: the assertion that these are superseded is categorically false. For any reasonably co
by physPop 10y ago
Largely superseded? This is a sarcastic comment, right?
In case it isn't: the assertion that these are superseded is categorically false. For any reasonably computationally difficult problem, being able to capture the structure of a problem is hugely powerful, rather than blindly throwing deep learning algorithms at it. Just becuase you can doesn't mean you should.
For example, algorithms for convex problems in particular can be orders of magnitude more efficient than naive nonlinear approaches. Also consider the case where the problem has some (possibly sparse) structure, where custom solvers can render trivial otherwise computationally intractable problems.
- scotty79 10y agoI read about application of neural networks to fluid dynamics. It ended up being faster than usual approaches. At least some matemathical solutions might eventually be superseeded by pretrained nn at least in some contexts.
- SmooL 10y agoI think the key was that the neural network didn't compute the fluid model accurately so much as a compute a simulation that looked accurate to humans
- srean 10y agoIndeed. Few people talk about were the deep (or for that matter shallow) dirty laundry is. NNs require a fantastic amount of babysitting and trying out different configurations, the first time around for a specific dataset/task. Once done, you do get great results.