Fujitsu

The Fujitsu Digital Annealer 🔗 provides an alternative to quantum computing technology

Solver

Model	Solvers
Quadratic Unconstrained Binary Optimization (QUBO)	fujitsu.DA3

Quadratic Unconstrained Binary Optimization (QUBO)

The Fujitsu solver can be used to solve QUBO problems written as a binary quadratic model (BQM).

import strangeworks as sw
from strangeworks_optimization import StrangeworksOptimizer
from strangeworks_optimization_models.parameter_models import (
    FujitsuParameterModel
)
from dimod import BinaryQuadraticModel

sw.authenticate(api_key)

linear = {1: -2, 2: -2, 3: -3, 4: -3, 5: -2}
quadratic = {(1, 2): 2, (1, 3): 2, (2, 4): 2, (3, 4): 2, (3, 5): 2, (4, 5): 2}
model = BinaryQuadraticModel(linear, quadratic, "BINARY")

options = FujitsuParameterModel(
        time_limit_sec=1, penalty_coef=10000
    )

solver = "fujitsu.DA3"

so = StrangeworksOptimizer(model=model, solver=solver, options=options)
sw_job = so.run()

print(f"Job slug: {sw_job.slug}")
print(f"Job status: {so.status(sw_job.slug)}")

results = so.results(sw_job.slug)

Fujitsu Model

Fujitsu API Reference 🔗

import strangeworks as sw
from strangeworks_optimization import StrangeworksOptimizer
from strangeworks_optimization_models.parameter_models import (
    FujitsuParameterModel
)
from strangeworks_optimization_models.problem_models import FujitsuModelList

sw.authenticate(api_key)

model = FujitsuModelList(
    binary_polynomial={
        "terms": [
            {"coefficient": 2, "polynomials": [1, 2]},
            {"coefficient": -4, "polynomials": [2, 4]},
            {"coefficient": -3, "polynomials": []},
        ]
    }
)

options = FujitsuParameterModel(
        time_limit_sec=1, penalty_coef=10000
    )

solver = "fujitsu.DA3"

so = StrangeworksOptimizer(model=model,
                            solver=solver,
                            options=options)
sw_job = so.run()

print(f"Job slug: {sw_job.slug}")
print(f"Job status: {so.status(sw_job.slug)}")

results = so.results(sw_job.slug)

Convert from QUBO

For larger models it is advisable to first convert the BinaryQuadraticModel to a FujitsuModelList locally before submitting.

import strangeworks as sw
from strangeworks_optimization import StrangeworksOptimizer
from strangeworks_optimization_models.problem_models import FujitsuModelList
from strangeworks_optimization_models.converter_models import StrangeworksConverterFactory

sw.authenticate(api_key)

linear = {1: -2, 2: -2, 3: -3, 4: -3, 5: -2}
quadratic = {(1, 2): 2, (1, 3): 2, (2, 4): 2, (3, 4): 2, (3, 5): 2, (4, 5): 2}
model = BinaryQuadraticModel(linear, quadratic, "BINARY", offset=1.4)
strangeworks_converter = StrangeworksConverterFactory.from_model(model, FujitsuModelList)

fujitsu_model = strangeworks_converter.convert()

optimizer = StrangeworksOptimizer(model=fujitsu_model, solver="fujitsu.DA3")
optimizer.run()

Parameters

from strangeworks_optimization_models.parameter_models import FujitsuParameterModel

options = FujitsuParameterModel(
        .....
    )

Solvers

• fujitsu.DA3

Parameter	Description	Type	Default	Range
time_limit_sec	Maximum running time of DA in seconds	int	10	1 to 3600
target_energy	Threshold energy for fast exit	float	disabled	-
num_run	The number of parallel attempts of each group	int	1	1 to 16
num_output_solution	The number of output solutions of each group	int	5	1 to 1024
gs_level	Level of the global search	int	5	0 to 100
gs_cutoff	Global search cutoff level	int	8000	0 to 1000000
one_hot_level	Level of the 1hot constraint search	int	3	0 to 100
one_hot_cutoff	Level of the convergence for 1hot constraint search	int	100	0 to 1000000
internal_penalty	Mode of 1hot constraint internal generation	int	0	0 or 1
penalty_auto_mode	Coefficient adjustment mode	int	1	0 to 10000
penalty_coef	Coefficient of the constraint term	int	1	1 to 9223372036854775807
penalty_inc_rate	Parameters for automatic adjustment of constraint terms	int	150	100 to 200
max_penalty_coef	Maximum constraint term coefficient	int	0	0 to 9223372036854775807
guidance_config	Initial value of each variable	dict	-	-
fixed_config	Fixed value of each variable	dict	-	-
one_way_one_hot_groups	Number of variables in each group of one-way one-hot constraints	dict	-	-
two_way_one_hot_groups	Number of variables in each group of two-way one-hot constraints	dict	-	-