Answered
You spin me right round... Why Ising?
Is there an application for the Ising model that would be relatable to non-physicists?
Other than analyzing spin matrices, what reasons might there be for using the Ising model instead of a QUBO?
Comments
Some graph optimization algorithms map more intuitively to Ising than QUBO. Examples would be Max-Cut or some satisfiability variants like Not-All-Equal 3-SAT.
Signed social networks also have a natural mapping to Ising.
If we only used the coupling terms in the two models then the Ising coupling creates a equal/not-equal relationship. Whereas the QUBO coupling creates an AND/NAND relationship. They each feature in different areas, and many times both are appropriate -- as in constraint satisfaction. The AND/NAND relationship is intuitive in applications of selection (e.g. selecting financial assets in portfolio optimization.)
That helps - thanks.
Your response got me looking for examples of algorithms using the Ising model. I found a paper by Andrew Lucas on arxiv.org, "Ising formulations of many NP problems." He frames Karp's 21 NP-complete problems using the Ising model. I don't know if the paper is referenced in any of the DWave literature, but at first pass it looks good.
https://arxiv.org/abs/1302.5843
Hello Thomas, we are very familiar with that paper. Many of the algorithms have been implemented on a D-Wave QPU. For example, we use his Traveling Salesman algorithm with the 48-city continental USA state capitals for a basic demo. If you implement some of them, and get interesting results, we'll be very interested.
Please sign in to leave a comment.