Talk:Quantum annealing

[untitled]
What is a glassy system? Can this term be changed to something that is defined or a definition page added? this seems like too important of a part of the quantum annealing definition to have an unlinked page... Pandemias 17:07, 8 March 2007 (UTC)
 * I think that link should point to Ising model. SuicidalCat (talk) 20:20, 27 June 2009 (UTC)


 * Actually the Ising model is a simplification for glassy systems that is easy to treat computationally. A glassy system is a more general concept that intuitively refers to a particle system where separate points in an n-dimensional space are ordered in respect to one another, following orientation rules from cristallography. For a reference, try spin glasses. This mathematical model can physically result in glass, gems, metal and many other inanimate matter aggregation states in Nature. The idea is that this natural basis is the substrate for a metaphor. This metaphor then in turn serves to construct a computable model to be used in "difficult" optimization tasks such as the "n-body problem". These problems usually require a lot of multidimensional integrals to be solved at once, and, being this very costly in computational terms, are approached through a stochastic strategy such as QA. This is because QA is much more efficient in large search spaces with high, but thin, interstate barriers. (Aswarp (talk) 17:41, 21 November 2011 (UTC)).

Pseudocode please, someone!
As in the simulated annealing article. Pseudocode is very helpful to those of us actually trying to code these algorithms. Thanks! --84.9.85.135 15:12, 31 October 2007 (UTC)


 * Quantum annealing is not just a mathematical optimization technique... it should be, first and foremost, a physical procedure, as thermal annealing is. Simulated annealing is a methematical technique, and simulated quantum annealing is also. Javirl (talk) 11:17, 3 July 2010 (UTC)


 * Source code and pseudocode can be found in the work of Robert S. Tucci and many others from the italian school, such as in "An introduction to quantum annealing" by Diego de Falco and Dario Tamascelli: http://arxiv.org/abs/1107.0794 (Aswarp (talk) 17:41, 21 November 2011 (UTC)).


 * The pseudocode would be that in the simulated annealing article modified with barrier avoidance. Normally this is done by taking a several random start points and following the SA algorithm on those (in parallel if you like), then comparing the final states. If you wanted to do it as though it were a single quantum state, you would have to have some idea of the entire energy landscape so that you can approximate the probability of tunneling effectively. Otherwise, just periodically test random points outside your current position and see if they give you something sufficiently good to try to run SA on for a bit. So those are the basic principles, and you can see the paper above for a more detailed model of the probabilities used. SamuelRiv (talk) 14:58, 6 July 2014 (UTC)

Scope of article
I'm worried this article is trying to describe two different but related things: Stephen J. Brooks (talk) 16:40, 18 December 2013 (UTC)
 * there's an algorithm on a normal (not quantum) computer, which is like the simulated annealing algorithm but changes step size rather than "temperature". This is related to ASA.
 * Then there's an algorithm on a quantum computer where there really is a wavefunction filling the potential wells of the function to be minimised. This process may also be simulated on a classical computer but the resulting algorithm is much slower than the one above.  On a quantum computer it's supposed to be faster and find the global optimum more often.

Problem with 1D energy diagram
I'm a little (maybe more than a little) concerned with the energy vs. configuration diagram. It's easy to swallow at first, but upon further consideration I suspect it is not an accurate representation. The problem is with the x-axis. What it suggests is that configurations must follow an exact sequence of changes to go from one form to the next, which I believe is false, even classically!

With n qubits (or bits for that matter), if I'm not mistaken, the configuration space should be something more like n-dimensional with 2 possible states in each dimension. It quickly becomes impossible to imagine the energy surface after 3 (qu)bits (though even with 3, imagining the surface is non-trivial).

For example, with 2 qutrits energy is a two dimensional surface with qutrit1 as the x-axis, and qutrit2 as the y-axis. There are only 3 states for each qutrit, so 32 total states. Let's say the energy surface is just a big hump in the middle. Let's also imagine we start at the origin, but the ground state is diagonally across from the hump. Classically, one trit would have to first change two states, and then the other would do the same; they can't simultaneously do this because of the energy hump. Here is where tunnelling wins because qutrits can tunnel through the hump.

- Michael W

174.112.30.70 (talk) 16:53, 25 November 2016 (UTC)

Incorrect Claim Regarding D-Wave?
There's the following line near the bottom of the article.

> D-Wave's architecture differs from traditional quantum computers. It is not known to be polynomially equivalent to a universal quantum computer and, in particular, cannot execute Shor's algorithm because Shor's algorithm is not a hillclimbing process.[Citation Needed]

However, as far as I understand, the Dwave machines are adiabatic? If that's the case it has been shown that the traditional circuit model and adiabatic process model are equivalent (up to a polynomial factor), per https://journals.aps.org/rmp/abstract/10.1103/RevModPhys.90.015002

If there is some detail I am missing here regarding implementation it should probably be documented!

(I apologize if I am misunderstanding something here) 66.168.46.68 (talk) 01:35, 19 December 2023 (UTC)


 * D-wave's machine is not a universal quantum computer as it cannot implement arbitrary unitaries. It is an annealer for an Ising spin glass on a particular graph (the Chimera graph, defined by the connectivity of the machine). The operative Hamiltonian is always diagonal in the z-basis. Moreover, in contrast to the adiabatic quantum computer, which assumes that on starts and remains in the ground state, the quantum annealer works at finite temperature. See some discussion of the latter in arXiv:1304.4595. --Qcomp (talk) 18:07, 19 December 2023 (UTC)