Talk:Quantum computer
|
|
| Missing image Cscr-featured.png Featured article star | Quantum computer is a featured article, which means it has been identified as one of the best articles produced by the Wikipedia community. If you see a way this page can be updated or improved without compromising previous work, feel free to contribute. |
"For problems with all four properties, it will take an average of n/2 guesses to find the answer using a classical computer. The time for a quantum computer to solve this will be proportional to the square root of n".
Shouldn't that be log(n)? If not, a little explanation on this rather unexpected outcome would be appreciated...
Apart from that it's a great piece of work! --MK
This quote in the article is nonsense:
"Integer factorization is believed to be practically impossible with an ordinary computer for large numbers (eg. 10^600)."
While that's a really big number, a high school freshman can factor that big number in his head! 2^600 x 5^600. Gee, that was tough!
- Thanks for spotting this. I changed the article accordingly. Of course, you could also have changed the article yourself. -- Jitse Niesen 15:06, 30 Jan 2005 (UTC)
- No problem. It sometimes seems easier to complain than to just fix it. I like your change (I couldn't come up with a suitable fix; you did it quite well.) I make a better critic than author. -- Seinfeld's_Cat 1 Feb 2005
The talk page is as interesting as the main article. Amazing! --ReallyNiceGuy
Great article!
Maybe we could add in the first paragraph that quantum computers are by their very nature probabilistic devices and only probabilistic algorithms can be implemented on them. Also, quantum computers can be simulated by Turing machines and are therefore no attack to the Church-Turing thesis. --AxelBoldt
I've now added both. They're in the complexity theory section, where they seemed to fit the flow best. -LC
Yes.
In the complexity section, it says that BQP is the class of problems that can be solved by quantum computers. These however are only the problems that can be solved by quantum computers in polynomial time. Maybe you could say "the problems that can be solved by quantum computers in reasonable time" or "that can be realistically solved by quantum computers".
Then it says that quantum computers probably cannot solve all NP-complete problems. There are two problems with this statement: strictly speaking, a quantum computer only works probabilistically and cannot "solve" any NP-complete problem (or any other decision problem for that matter) in the same sense a Turing machine solves them, with a deterministic and correct Yes/No answer. Furthermore, if we allow probabilistic solutions, then quantum computers can of course solve all NP-complete problems, just like any Turing machine can; it may just take a lot of time to do so... --AxelBoldt
It should be clearer now. -LC
What is this #P-Complete ? --Taw
See #P-Complete. -LC
How long does it take to factor an n-digit number with n qubits? --Axel
Would Quantum Computing break Elliptic Curve Cryptography ? Taw
- Apparently so, through a variation on Shor's algorithm. I haven't studied it, though -- CYD
Some recent work [1] (http://www.eet.com/story/OEG20010924S0101) indicates that if spaced sufficiently closely, quntum entanglement between quantum dots may be possible, so it's possible that in the future a quantum computer could be implemented using quantum dots.
Also, it might be useful to mention "reversibility", the haddamard(sp?) transformation and the various types of quantum logic gates (CNOT, etc...). I'm not an expert, so I'll defer to someone who knows about this stuff -- Matt Stoker
What does it mean to have a "quantum computer with nonlinear operators"? --Robert Merkel
I hope that's clearer now. --LC
How the heck do you implement a nonlinear operation on qubits? Evolution always proceeds by unitary - therefore linear - operations. -- CYD
I agree, that seems to be a consequence of Schrodinger's equation, which is pretty basic to QM. --AxelBoldt
I agree. The papers proving the result never said it could be done. But it's an interesting result. It hasn't been ruled out that a large linear system could act like a small nonlinear system, and give the desired result. The Shi paper refers to the nonlinearity of the Gross-Pitaevskii equations, but I'm not familiar with them. I would assume the Shi paper is flawed, since it wasn't accepted anywhere, but I'm not aware of any proofs that this sort of thing is inherently impossible. --LC
Oh. I just checked the site, and the Shi paper has now been accepted in a journal. It hadn't yet been accepted when I first wrote the article. I'll remove the "not peer reviewed" note. Is anyone here familiar with the "Internation Journal of Modern Physics"? Is it generally respectable? --LC
- The quantum computer in the above example could be thought of as a black box containing 8 complex numbers. Or, it could be thought of as 8 black boxes, each containing 1 complex number, each sitting in a different alternate universe, and all communicating with each other. These two interpretations correspond to the Copenhagen interpretation and many worlds interpretation, respectively, of quantum mechanics. The choice of interpretation has no effect on the math, or on the behavior of the quantum computer. Either way, it's an 8-element vector that is modified by matrix multiplication
I don't think this characterization of many worlds is correct. The different universes don't come with complex numbers attached. Instead, the more likely states are exhibited in more universes. The goal of the matrix manipulations is to bring essentially all universes into the same state. Once you measure the state, the universes stop to communicate and truly split. --AxelBoldt
Good point. I've reworded it. --LC
In the "How they work" section it says:
"The square of (a+ib) is (a^2+b^2)."
Actually the square of (a+ib) would be (a^2-b^2) + i(2ab).
Probably you mean the norm or the mod or something like that? (it's been a while since i used complex numbers, not sure of the name any more).
In quantum mechanics, (a+bi) is called the amplitude, and (a2+b2) is called the squared amplitude, even though the latter equals |a+bi|2 rather than the (a+bi)2 that the name might suggest. I suppose the terminology is counterintuitive. I'll reword the article to make it clearer. --LC
I've removed the following contribution from User:Harry Potter:
- Quantum computing can also theoretically produce time anomalies. The ability of the Quantum computer to find the correct answer in a factorisation problem can be seen as running all the possibilities simultaneously. However, only the correct answer produces a positive response which is then sent back through time to become the only calculation which in fact needs to be made.
Quantum computation doesn't involve time travel in any description I've seen. -- Oliver P. 00:01 9 Jun 2003 (UTC)
- Agreed. This sounds purely abstract and theoretical to me, but I am no expert on quantum computing by any means. Perhaps Harry Potter can produce a source in the quantum computing literature that lends some support to this idea? -- Wapcaplet 00:10 9 Jun 2003 (UTC)
Please specify "... have recently been built" into "... have been built in the early/mid-/late 20th century/21st century". --Menchi 08:37 14 Jul 2003 (UTC)
I found the first paragraph of the second section, titled "The Power of Quantum Computers", rather unclear. Unfortunately, I cannot tell exactly what makes it unclear - maybe it is too vague; how difficult is factorization (suspected to be) on a classical computer, and what is the speed-up on a quantum computer? I do not know enough about quantum computing to rewrite the paragraph myself ...
I liked the rest of the article very much, it contains (almost) everything that I was looking for. However, I am still wondering whether there is an upper bound on the power of quantum computers. The article says that quantum computers cannot solve undecidable problems, but is BQP known/suspected to be a subset of another complexity class, say NP? -- Jitse Niesen 11:08, 26 Aug 2003 (UTC)
| Contents |
They CAN solve Turing-unsolvable problems, but...
This is a technicality that should be mentioned: If you define a quantum computer as a Turing Machine with arbitrary complex amplitudes, then the class of machines that you obtain is uncountably infinite, and can easily be shown [ADH97] to contain machines that solve all kinds of Turing-unsolvable problems. But *we* can't build those QC's (or even know one if we see one) because of our inability to know the correct amplitudes required for those tasks. The computability comments in the present version of this article are valid for QC's whose amplitudes are computable numbers.
[ADH97] L. Adleman, J. DeMarrais, and M. Huang. Quantum computability, SIAM Journal on Computing 26:1524-1540, 1997.
Key exchange versus encryption
You say:
"This ability would allow a quantum computer to break many of the cryptographic systems in use today. In particular, most of the popular public key ciphers could be quickly broken, including forms of RSA, ElGamal and Diffie-Hellman."
My understanding is that RSA is the encyrption method and the ElGamal and Diffie-Hellman are key exchange protocols. The references within Wikipedia also say that. Do you mean that the encyrpted key being transmitted is able to be read? BK
Further editing
Some of the section which was there before (and is now called Initialization and termination), though written at a good level I believe contains some innaccuracies, i.e. on how the qubit registers are measured. In fact, only a fixed register is sampled on termination.CSTAR 04:27, 6 Jun 2004 (UTC)
Oops that's not what I should have said. Anyway I think it's right now. I also split the article up adding the stuff on reversible registers to quantum circuits.CSTAR 07:01, 6 Jun 2004 (UTC)
Where did the numbers come from in columns #2 and #3 in the "bits vs qubits" section?
Why is the 3-qubit "000" represented as "0.37 + i 0.04"? Where did the 0.37 and 0.04 come from? Ditto for the rest of the numbers in that table. According to www.qubit.org, a 3-qubit system can represent 8 values. There is no magic in 16 analog values vs 8 complex values, there are just 8 values total. None of the references I have come across mention anything other than an L qubit system representing exactly 2^L possibilities at once, not 2^(L+1) (the 16 you mentioned in this section).
In an effort to give the reader some sense of how probability waves can be represented by complex numbers and how complex numbers c1an be manipulated, you've obfuscated the simpler facts of qubits. 1 qubit holds 2 states simultaneously with *equal probability*, 2 qubits hold 4 states simultaneously with *equal probability* and so on.
- I didn't write this example, but as an example, I think it's mostly OK (But you're right about the dimension being wrong: see below. When I read it I wasn't thinking) . What the example is trying to express is that fact that in a quantum computer the set of registers is a quantum mechanical object,and the state is described by a wave funbction on it. As far as the number of independent values, the example is meant to suggest the real dimension of the space of states, but you may be right the dimension counting may be wrong, because the relevant space is complex projective space of dimension 2n. Thanks. I'll fix it. The set of pure sets is homeomorphic to the compact symmetric space with m = 2n
- <math> \mathbf{U}(m)/ \mathbf{U}(m-1) \times \mathbf{U}(1) <math>
- As far as your statement:
- 1 qubit holds 2 states simultaneously with *equal probability*, 2 qubits hold 4 states simultaneously with *equal probability* and so on.
- I don't believe this statement is correct.
CSTAR 03:10, 16 Jun 2004 (UTC)
P.S. From this log, it wasn't clear who wrote the question and who wrote the reponse, so I went back and edited the past, by separating my answer to the anonymous question. I'm perfectly happy to edit the past. More of us should try it.CSTAR 04:48, 16 Jun 2004 (UTC)
other confusing bits
(problem I noted has been fixed -- thanks, everyone. -- DavidCary 01:43, 4 Jul 2004 (UTC))
- Yes, it's rubbish, I rewrote the section. Unfortunately the bit about fault tolerant scaling in Shor's algorithm is a bit hazy, I'm just quoting vague unpublished numbers floating around my research group. -- Tim Starling 07:35, Jun 18, 2004 (UTC)
- Good rewrite! Regarding range for T2, I checked Nielsen & Chuang book (pp. 337) and they claim 7 seconds and 0.3 seconds for proton and carbon in a two-qubit experiment. Should we change "hundreds of microseconds" to "seconds"? Or is there some caveat about Nielsen & Chuang data that I'm not aware of? Andris 18:37, Jun 18, 2004 (UTC)
- Alright. It would be nice to explain at some stage why systems such as chloroform nuclear spins are not scalable, but I guess "nanoseconds to seconds" will do for now. -- Tim Starling 01:47, Jun 22, 2004 (UTC)
The article currently mentions the transverse relaxation time T2 (terminology used in NMR and MRI technology)
Is there really any difference between NMR quantum computers and MRI quantum computers ? I suspect they're really the same thing. -- DavidCary 01:43, 4 Jul 2004 (UTC)
- I guess my comments weren't clear -- I was referring to MRI and NMR technology not computers (MRI technology, the kind used in hospitals) which are essentially the same technology (I think at least they are, since the both involve involve nuclear spin), but seem to have separate wiki articles. Is my making these two references too confusing? Please change it if you can make it clearer. CSTAR 01:51, 4 Jul 2004 (UTC)
Interpretation
- Now that we're at it, I think we should remove references to interprertation (Copenhagen, many worlds etc..). Regardless of content, they shouldn't be in this article. I haven't made an effort to understand whether what's writtem in the article makes any sense, so I won't pass judgment on that account_. CSTAR 13:19, 18 Jun 2004 (UTC)
Edits of 08/01/2004
Some of the edits by an anonymous user did increase readability, but I think calling entanglement or superposition a process is misleading. Also Qubits don't hold data -- they are measures of the size of data. The register space is the actual container of data. I tried to keep the spirit of the recent edits, but modifying what seemed misleading. CSTAR 02:49, 2 Aug 2004 (UTC)
---
So, how the hell are you supposed to make software for these? I assume a normal "if something is something do this, else do this" doesn't work?
- Conditionals in the usual sense, no.CSTAR 20:08, 16 Aug 2004 (UTC)
Decoherence and Complexity
This article is really great !
Nevertheless, there is one point I found unclear : the relationship between decoherence and complexity.
The article state some algortihms are in BQP. Since complexity classes are only meaningful for arbitrarily large calculations, does the definition of the BQP class includes the fact that decoherence may limits the amount of operations that may be performed, or does it assumes we have a way to prevent the effects of decoherence for arbitrarily long calculations (which doesn't seem obvious) ?
For example, an old (1995) article by P. Shor and coll. "Quantum Computers, Factoring and Decoherence" (http://citeseer.ist.psu.edu/chuang95quantum.html), calculated a limit to the size of the integers that may be factored faster by a quantum computer than by a classic computer, which to me would contradict the fact that factoring is in BQP.
Some clarification of this point would be great !
Thanks !
- That's a good question, but I'm inclined to say that the there is no relationship othre than that which is buit into the model itself. To explain rigorously BQP, you have to formulate a more precise model of quantum computation than is done in this article. The right model is that of quantum circuit, although BQP is not defined there either, although it should be. At some point I'll get around to it. Otherwise look at some of the references such as Kitaev, or Nielsen Chuang. The book by Hirvensalo is also good. I know I have evaded your question but that's the best I can do. Maybe some other wikipedian has a better answer.CSTAR 13:57, 1 Oct 2004 (UTC)
- The definition of BQP assumes that computation is error-free (and decoherence-free), which is not the case in practice. The limit in Shor's 1995 article is for quantum computers which are subject to decoherence. Thus, there is no contradiction: the 1995 paper essentially says that if there was not decoherence, quantum computers could factor large numbers, but in presence of decoherence, they could not. This is the (pessimistic) view of quantum computation that was common in mid-1990s. This view is mostly obsolete now. Quantum error correction has been invented (in a later work of Shor and colleagues in 1996 and 1997) and it has been shown that arbitrarily large computations can be made fault-tolerant (with "fault" including "decoherence"), as long as fault rate is sufficiently small. Andris 00:10, Dec 11, 2004 (UTC)
Cost of quantum computations
The big selling point of quantum computations seems to be that quantum computers with enough qubits can run algorithms much faster than classicalcomputers. That's great, but it's not very useful if, for example, the cost of implementing an n-qubit register is exponential in n. Is there some estimate for the asymptotics of this? Is the apparent advantage of quantum computers due only to the fact that we're measuring computational complexity by the wrong yardstick? (I have no idea, but someone surely knows...) --Andrew 08:48, Dec 7, 2004 (UTC)
NP-Complete stuff
"BQP is suspected to be disjoint from NP-complete and a strict superset of P, but that is not known. Both integer factorization and discrete log are in BQP. Both of these problems are NP problems suspected to be outside P. Both are suspected to not be NP-complete. There is a common misconception that quantum computers can solve NP-complete problems in polynomial time. That is not known to be true, and is generally suspected to be false."
A teacher claimed that someone has found an quantum algorithm to solve NP-complete problems in polynomial time. Asking for a size estimate, he said that the algorithm description should approximately fit on a page. He's supposed to be teaching everyone the definition of NP-complete problems next week, so he should know what he's talking about. He seemed rather sure. I'd like to see that algorithm... (Haven't had time to search the internet yet.) Would sure make quantum computers (even) more interesting... Κσυπ Cyp 17:27, 9 Dec 2004 (UTC)
- Regarding the previous quote: Are integer factorization and discrete log (suspected to be) outside BPP? I think this is more interesting than whether they are outside P. -- Jitse Niesen 19:05, 9 Dec 2004 (UTC)
- Yes, they are thought to be outside BPP. Andris 04:32, Dec 10, 2004 (UTC)
- Thanks, I updated the article. -- Jitse Niesen 13:42, 10 Dec 2004 (UTC)
Clustered quantum computer ?
Hello,
I don't understand a lot about the technics of quantum computers. I would like to know if it is possible (at least in theory) to distributed quantum computing. it is possible to dream about a “quantum grid”, “quantum Beowulf”, “quantum@home?”
Thank you
? David Latapie 07:31, 25 Jan 2005 (UTC)
Can quantum computers solve undecidable problems?
This article says that quantum computers cannot solve any problems that classical computers cannot solve either. However, after a recent edit, the start of Matiyasevich's theorem reads:
- "Matiyasevich's theorem, proven in 1970 by Yuri Matiyasevich, implies that Hilbert's tenth problem is unsolvable (although it has recently been proposed by Tien Kieu [Int.J.Theor.Phys. 42 (2003) 1461-1478] that this result only applies to a deterministic Turing machine, and that the problem could be solved on a quantum computer using a quantum adiabatic algorithm. This is still contraversial.)."
How to reconcile both quotes? Does anybody know how contraversial Tien Kieu's work is? -- Jitse Niesen 13:36, 10 Feb 2005 (UTC)
- Probabilistic Turing machines or non-deterministic Turing machines don't decide more problems than plain old Turing machines. Neither does the quantum circuit model (and equivalents). Unfortunately, I don't know enough what a quantum adiabatic algorithm is but it possibly doesn't fall under the quantum circuit model of quantum computation. I suppose I should look at the paperCSTAR 14:41, 10 Feb 2005 (UTC)
- I just looked at one of his papers. It is highly suspect, making claims about probabilistic computation which under any normal interpretation are false. "Probabilistic computation may be more powerful than Turing Computation provided a suitable probability measure can be defined." Now it is well-known, (see Sipster's textbook on computation) that probabilistic Turing machines compute the same functions as normal ones.
- I would take these claims with lots of salt (and maybe that sentence in Matiyasevich's theorem should be removed).CSTAR 18:15, 10 Feb 2005 (UTC)
Thank you; I came to the same conclusion. -- Jitse Niesen 21:47, 10 Feb 2005 (UTC)
Request for references
Hi, I am working to encourage implementation of the goals of the Wikipedia:Verifiability policy. Part of that is to make sure articles cite their sources. This is particularly important for featured articles, since they are a prominent part of Wikipedia. Now this article has an extensive list of additional material, but list of further reading is not the same thing as proper references. Further reading could list works about the topic that were not ever consulted by the page authors. If some of the works listed in the that section were used to add or check material in the article, please list them in a references section instead. The Fact and Reference Check Project has more information. Thank you, and please leave me a message (http://en.wikipedia.org/w/wiki.phtml?title=User_talk:Taxman&action=edit§ion=new) when a few references have been added to the article. - Taxman 19:18, Apr 22, 2005 (UTC)
Factorization
"For problems with all four properties, it will take an average of n/2 guesses to find the answer using a classical computer. The time for a quantum computer to solve this will be proportional to the square root of n...But it is also easy to defend against. Just double the size of the key for the cipher."
This doesn't sound right to me. If your key length is 2n, a classical computer would take n guesses. A quantum computer would then take time proportional to sqrt(2n), which is asymptotically the same as sqrt(n).
I didn't edit the page because I'm not an expert on quantum computing, and feel it's possible that I don't understand (if, for example, you can't do asymptotic analysis on quantium computing algorithms?). As best as I can figure, though, I would expect you'd need to square the key length?
Any problem with these four problems will take an average of (n+1)/2 guesses to find the answer: Fix the guesses you are going to try. Each guess then has probability 1/n of being correct. Then expected guesses is <math>\sum_{i=1}^n i \cdot \frac{1}{n} = \frac{n(n+1)}{2n} = \frac{n+1}{2}<math>.
- n here is the size of the search space, which for a brute force attack on a symmetric cipher is equal to 2L where L is key length. So the classical algorithm takes O(2L) and the quantum algorithm takes O(2L/2). Note that this applies to Grover's algorithm, which is not the usual algorithm used for factorisation. It's a slower algorithm for a more general class of problems. -- Tim Starling 19:22, Jun 12, 2005 (UTC)