Why Do Results Vary?

You can run the same problem on the QPU more than once and get different results.

The DWaveSampler uses the D-Wave™ QPU as a computational resource. QPUs naturally return low-energy solutions, but they are inherently probabilistic.

For example, you can imagine sampling from a single qubit with no bias and no interactions with other qubits:

sampler = DWaveSampler()
resp = sampler.sample_ising({0: 0.0}, {}, num_reads=1)

Everything else being equal, the probability of the qubit ending in the -1 or the 1 state is equal (50 percent).

In another example, say we strongly bias a single qubit towards -1:

sampler = DWaveSampler()
resp = sampler.sample_ising({0: 2.0}, {})

Even in this case, there is a small chance that the QPU will end the anneal in an excited state and return 1, rather than -1.

It is also important to note that because this sampling is done via a physical process, there is no 'state' to fix.