5. Solvers¶

5.1. LearningSolver¶

Example¶

[3]:

import random

import numpy as np
from scipy.stats import uniform, randint
from sklearn.linear_model import LogisticRegression

from miplearn.classifiers.minprob import MinProbabilityClassifier
from miplearn.classifiers.singleclass import SingleClassFix
from miplearn.collectors.basic import BasicCollector
from miplearn.components.primal.actions import SetWarmStart
from miplearn.components.primal.indep import IndependentVarsPrimalComponent
from miplearn.extractors.AlvLouWeh2017 import AlvLouWeh2017Extractor
from miplearn.io import save
from miplearn.problems.tsp import (
    TravelingSalesmanGenerator,
    build_tsp_model,
)
from miplearn.solvers.learning import LearningSolver

# Set random seed to make example reproducible.
random.seed(42)
np.random.seed(42)

# Generate a few instances of the traveling salesman problem.
data = TravelingSalesmanGenerator(
    n=randint(low=10, high=11),
    x=uniform(loc=0.0, scale=1000.0),
    y=uniform(loc=0.0, scale=1000.0),
    gamma=uniform(loc=0.90, scale=0.20),
    fix_cities=True,
    round=True,
).generate(50)

# Save instance data to data/tsp/00000.pkl.gz, data/tsp/00001.pkl.gz, ...
all_data = save(data, "data/tsp")

# Split train/test data
train_data = all_data[:40]
test_data = all_data[40:]

# Collect training data
bc = BasicCollector(time_limit_sec=3600)
bc.collect(train_data, build_tsp_model, n_jobs=4)

# Build learning solver
solver = LearningSolver(
    components=[
        IndependentVarsPrimalComponent(
            base_clf=SingleClassFix(
                MinProbabilityClassifier(
                    base_clf=LogisticRegression(),
                    thresholds=[0.95, 0.95],
                ),
            ),
            extractor=AlvLouWeh2017Extractor(),
            action=SetWarmStart(),
        )
    ]
)

# Train ML models
solver.fit(train_data)

# Solve a test instance
solver.optimize(test_data[0], build_tsp_model);

Gurobi Optimizer version 9.5.2 build v9.5.2rc0 (linux64)
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 10 rows, 45 columns and 90 nonzeros
Model fingerprint: 0x6ddcd141
Coefficient statistics:
  Matrix range     [1e+00, 1e+00]
  Objective range  [4e+01, 1e+03]
  Bounds range     [1e+00, 1e+00]
  RHS range        [2e+00, 2e+00]
Presolve time: 0.00s
Presolved: 10 rows, 45 columns, 90 nonzeros

Iteration    Objective       Primal Inf.    Dual Inf.      Time
       0    6.3600000e+02   1.700000e+01   0.000000e+00      0s
      15    2.7610000e+03   0.000000e+00   0.000000e+00      0s

Solved in 15 iterations and 0.01 seconds (0.00 work units)
Optimal objective  2.761000000e+03
Set parameter LazyConstraints to value 1
Gurobi Optimizer version 9.5.2 build v9.5.2rc0 (linux64)
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 10 rows, 45 columns and 90 nonzeros
Model fingerprint: 0x74ca3d0a
Variable types: 0 continuous, 45 integer (45 binary)
Coefficient statistics:
  Matrix range     [1e+00, 1e+00]
  Objective range  [4e+01, 1e+03]
  Bounds range     [1e+00, 1e+00]
  RHS range        [2e+00, 2e+00]

User MIP start produced solution with objective 2796 (0.01s)
Loaded user MIP start with objective 2796

Presolve time: 0.00s
Presolved: 10 rows, 45 columns, 90 nonzeros
Variable types: 0 continuous, 45 integer (45 binary)

Root relaxation: objective 2.761000e+03, 14 iterations, 0.00 seconds (0.00 work units)

    Nodes    |    Current Node    |     Objective Bounds      |     Work
 Expl Unexpl |  Obj  Depth IntInf | Incumbent    BestBd   Gap | It/Node Time

     0     0     cutoff    0      2796.00000 2796.00000  0.00%     -    0s

Cutting planes:
  Lazy constraints: 3

Explored 1 nodes (15 simplex iterations) in 0.02 seconds (0.00 work units)
Thread count was 12 (of 12 available processors)

Solution count 1: 2796

Optimal solution found (tolerance 1.00e-04)
Best objective 2.796000000000e+03, best bound 2.796000000000e+03, gap 0.0000%

User-callback calls 103, time in user-callback 0.00 sec
Gurobi Optimizer version 9.5.2 build v9.5.2rc0 (linux64)
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 10 rows, 45 columns and 90 nonzeros
Model fingerprint: 0x74ca3d0a
Variable types: 0 continuous, 45 integer (45 binary)
Coefficient statistics:
  Matrix range     [1e+00, 1e+00]
  Objective range  [4e+01, 1e+03]
  Bounds range     [1e+00, 1e+00]
  RHS range        [2e+00, 2e+00]
Presolved: 10 rows, 45 columns, 90 nonzeros

Continuing optimization...


Cutting planes:
  Lazy constraints: 3

Explored 1 nodes (15 simplex iterations) in 0.01 seconds (0.00 work units)
Thread count was 12 (of 12 available processors)

Solution count 1: 2796

Optimal solution found (tolerance 1.00e-04)
Best objective 2.796000000000e+03, best bound 2.796000000000e+03, gap 0.0000%

User-callback calls 27, time in user-callback 0.00 sec

[2]:

4. Primal Components 6. Benchmark Problems