Attention

dwave-greedy is deprecated since dwave-ocean-sdk 6.1.0 in favor of dwave-samplers and will be removed in dwave-ocean-sdk 8.0.0.

Samplers#

The dwave-greedy package currently includes just one sampler, SteepestDescentSampler, which is an alias for SteepestDescentSolver.

A sampler accepts a binary quadratic model (BQM) and returns variable assignments. Samplers generally try to find minimizing values but can also sample from distributions defined by the BQM.

SteepestDescentSolver#

Class#

class SteepestDescentSolver[source]#

Steepest descent sampler for binary quadratic models.

Steepest descent is the discrete analogue of gradient descent, but the best move is computed using a local minimization rather rather than computing a gradient. The dimension along which to descend is determined, at each step, by the variable flip that causes the greatest reduction in energy.

Solves convex problems to optimality.

Number of downhill runs (samples produced) is determined by num_reads, number of initial_states, or a combination of the two, depending on the initial_states_generator.

For a given input model’s graph \(G = (V, E)\), \(V\) being a set of graph vertices and \(E\) a set of edges, runtime complexity of the underlying C++ implementation is \(O(|E|)\) for initialization phase and \(O(|V|)\) per downhill step.

In the large_sparse_opt mode, runtime complexity on sparse graphs is \(O(|V|*log|V|)\) for initialization and \(O(max\_degree * log|V|)\) per downhill step.

Aliased as SteepestDescentSampler.

Examples

Solve a simple Ising problem.

>>> from dwave.samplers import SteepestDescentSampler
...
>>> sampler = SteepestDescentSampler()
>>> samples = sampler.sample_ising({0: 2, 1: 2}, {(0, 1): -1})
...
>>> print(samples)      
   0  1 energy num_oc. num_st.
0 -1 -1   -5.0       1       2
['SPIN', 1 rows, 1 samples, 2 variables]

Post-processes samples generated by another sampler (simulated annealing in this example):

>>> from dwave.samplers import SteepestDescentSampler, SimulatedAnnealingSampler
>>> import dimod
...
>>> bqm = dimod.generators.ran_r(5, 3)
>>> samples = SimulatedAnnealingSampler().sample(bqm)
>>> postprocessed = SteepestDescentSampler().sample(bqm, initial_states=samples)

For additional examples, see sample().

Attributes#

`SteepestDescentSolver.properties`	Values for parameters accepted by the sampling methods.
`SteepestDescentSolver.parameters`	Keyword arguments accepted by the sampling methods.

Methods#

`SteepestDescentSolver.sample`(bqm[, ...])	Find minima of a binary quadratic model.
`SteepestDescentSolver.sample_ising`(h, J, ...)	Sample from an Ising model using the implemented sample method.
`SteepestDescentSolver.sample_qubo`(Q, ...)	Sample from a QUBO using the implemented sample method.

SteepestDescentSampler#

Class#

SteepestDescentSampler[source]#: alias of SteepestDescentSolver