6 ms·
Qudits: The Real Future of Quantum Computing?
- petersjt014 9y agoI know these guys understand quantum anything more than I ever will, but is having a larger radix specifically a practical way for increasing performance? If a binary device has a hardware fault, bits that are in error still have a 50% chance of being correct. It seems like a single component failing here would make this worse (five times worse if it wasn't quantum). I don't know whether or not the pros outweigh the cons here, or at what radix they do (there's probably some statistical analysis that can answer that), but ten just _seems_ like too much.
- jdaaph 9y agoquantum algorithms change complexity class. The pro is simply too big if it can work in reality.
- gaze 9y agoA qudit expands the state space, sure. But now you need universal control over this system so I'm not sure you've bought anything. Also, a system will decay from energy level n to n-1 faster when n is larger... so like eh.
- gizmo686 9y agoI am not an expert. The issue that quantum computers have is not so much that individual qubits get errors, but rather that quibits become entangled with the rest of the universe, erasing our ability to take advantage of any quantom effect. To prevent this, you need to very carefully contain the quibits to prevent them from interacting with anything apart from highly controlled manipulations. The more quibits you have, the harder this becomes. If you can replace this with quidits that have more states, you can do the same computation, while needing fewer particles, making containment easier.
- adamnemecek 9y agoYou don't actually want qubits, you want an analog computer with differentiable signals. Most likely photonic. Qubits are a dead evolution branch. I've been recently exploring computational metamaterials for photonic computation. http://users.ece.utexas.edu/~aalu/research%20-%20page%203.htm http://users.ece.utexas.edu/~aalu/research%20-%20page%203.ht... (there's quite a few papers on this but unfortunately they are all paywalled. Spoiler alert, they seem to be based entirely on Fourier transform). These computational metamaterials don't need electricity to be powered (you need something that will shoot the photons on them and read back the values off tho). Machine learning would be much, much faster on these as you have O(1) differential calculus. They don't heat up. You can possibly build a house sized CPU out of these. I can see it, a city block sized CPU and a nuclear reactor next to it. Did you know that on an analog machine, you can do sort in O(n)? https://en.wikipedia.org/wiki/Spaghetti_sort https://en.wikipedia.org/wiki/Spaghetti_sort Hit me up if you wanna chat about this. I've seen the "light" (xdddd) now and can't go back to stupid bits. I'm not like super married to the metamaterials but analog photonic trumps quantum for just about every task I can think of.
- jd007 9y ago> You don't actually want qubits, you want an analog computer with differentiable signals. Most likely photonic. Qubits are a dead evolution branch. that's quite a bold claim. im pretty intrigued, but it seems like there isn't a lot of easily accessible information on this topic. the wikipedia article on analog computers mostly focuses on mechanical ones from the past. do you have any recommended starter readings for analog computation? is it a model that has the computational time complexity of a non-deterministic turing machine? thanks.
- adamnemecek 9y agoThere really isn't. Some of the old stuff still applies. But my approach to understanding this is very all over the place and I wouldn't recommend it for anyone besides myself. But I've found trying to understand electromagnetism and basic quantum theory (photon-matter interactions) to be quite helpful. As for classification, there's some work on this e.g. this http://www.cs.princeton.edu/~ken/MCS86.pdf http://www.cs.princeton.edu/~ken/MCS86.pdf I believe it's not non-deterministic, as in some sense, if you can pose the problem in a differentiable way, it's better than non-determinism. It's like being in a maze. An analog computer like robot can perceive the slight tilt of the floor that goes from the entrance to the exit and just follows that tilt. It takes the robot an optimal number of steps too. A non-deterministic Turing would kind of split itself into N different robots and start exploring all the turns. It might find the answer eventually but e.g. it will consume that much more energy. This analogy might be silly but I hope it conveys how very different analog machines are.
- cortexio 9y agoHow about.. oh i don't know... actually building a quantum computer?? no?? ok. Sry..