Files
ortools-clone/ortools/pdlp/iteration_stats_test.cc
2025-09-29 17:21:58 +02:00

736 lines
31 KiB
C++

// Copyright 2010-2025 Google LLC
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "ortools/pdlp/iteration_stats.h"
#include <cmath>
#include <limits>
#include <optional>
#include <utility>
#include "Eigen/Core"
#include "gtest/gtest.h"
#include "ortools/base/gmock.h"
#include "ortools/base/parse_text_proto.h"
#include "ortools/pdlp/quadratic_program.h"
#include "ortools/pdlp/sharded_quadratic_program.h"
#include "ortools/pdlp/solve_log.pb.h"
#include "ortools/pdlp/solvers.pb.h"
#include "ortools/pdlp/test_util.h"
namespace operations_research::pdlp {
namespace {
using ::google::protobuf::contrib::parse_proto::ParseTextOrDie;
using ::testing::AllOf;
using ::testing::Each;
using ::testing::ElementsAre;
using ::testing::Eq;
using ::testing::EqualsProto;
using ::testing::Ge;
using ::testing::Le;
using ::testing::Ne;
using ::testing::SizeIs;
using ::testing::proto::Approximately;
using ::testing::proto::Partially;
// The following block relies heavily on `EqualsProto`, which isn't open source.
void CheckScaledAndUnscaledConvergenceInformation(
QuadraticProgram qp, const Eigen::VectorXd& primal_solution,
const Eigen::VectorXd& dual_solution,
const double componentwise_primal_residual_offset,
const double componentwise_dual_residual_offset,
const ConvergenceInformation& expected_stats) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(std::move(qp), num_threads, num_shards);
EXPECT_THAT(
ComputeScaledConvergenceInformation(
PrimalDualHybridGradientParams(), sharded_qp, primal_solution,
dual_solution, componentwise_primal_residual_offset,
componentwise_dual_residual_offset, POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(expected_stats))));
// Rescale the problem so that the primal and dual solutions have elements
// equal to -1.0, 0.0, or 1.0.
Eigen::VectorXd col_scaling_vec = primal_solution.unaryExpr(
[](double x) { return x != 0.0 ? std::abs(x) : 1.0; });
Eigen::VectorXd row_scaling_vec = dual_solution.unaryExpr(
[](double x) { return x != 0.0 ? std::abs(x) : 1.0; });
Eigen::VectorXd scaled_primal_solution =
primal_solution.cwiseQuotient(col_scaling_vec);
Eigen::VectorXd scaled_dual_solution =
dual_solution.cwiseQuotient(row_scaling_vec);
sharded_qp.RescaleQuadraticProgram(col_scaling_vec, row_scaling_vec);
EXPECT_THAT(
ComputeConvergenceInformation(
PrimalDualHybridGradientParams(), sharded_qp, col_scaling_vec,
row_scaling_vec, scaled_primal_solution, scaled_dual_solution,
componentwise_primal_residual_offset,
componentwise_dual_residual_offset, POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(expected_stats))));
// Also check that the iteration stats for the scaled problem have the correct
// objectives and norms.
ConvergenceInformation expected_scaled_stats;
expected_scaled_stats.set_primal_objective(expected_stats.primal_objective());
expected_scaled_stats.set_dual_objective(expected_stats.dual_objective());
expected_scaled_stats.set_l_inf_primal_variable(1.0);
expected_scaled_stats.set_l_inf_dual_variable(1.0);
EXPECT_THAT(
ComputeScaledConvergenceInformation(
PrimalDualHybridGradientParams(), sharded_qp, scaled_primal_solution,
scaled_dual_solution, componentwise_primal_residual_offset,
componentwise_dual_residual_offset, POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(expected_scaled_stats))));
}
void CheckScaledAndUnscaledInfeasibilityStats(
QuadraticProgram qp, const Eigen::VectorXd& primal_ray,
const Eigen::VectorXd& dual_ray,
const Eigen::VectorXd& primal_solution_for_residual_tests,
const InfeasibilityInformation& expected_infeasibility_info) {
const int num_threads = 2;
const int num_shards = 2;
ShardedQuadraticProgram sharded_qp(std::move(qp), num_threads, num_shards);
EXPECT_THAT(
ComputeInfeasibilityInformation(
PrimalDualHybridGradientParams(), sharded_qp,
Eigen::VectorXd::Ones(sharded_qp.PrimalSize()),
Eigen::VectorXd::Ones(sharded_qp.DualSize()), primal_ray, dual_ray,
primal_solution_for_residual_tests, POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(expected_infeasibility_info))));
// Rescale the problem so that the primal and dual certificates have elements
// equal to -1.0, 0.0, or 1.0.
Eigen::VectorXd col_scaling_vec = primal_ray.unaryExpr(
[](double x) { return x != 0.0 ? std::abs(x) : 1.0; });
Eigen::VectorXd row_scaling_vec =
dual_ray.unaryExpr([](double x) { return x != 0.0 ? std::abs(x) : 1.0; });
Eigen::VectorXd scaled_primal_solution =
primal_ray.cwiseQuotient(col_scaling_vec);
Eigen::VectorXd scaled_dual_solution =
dual_ray.cwiseQuotient(row_scaling_vec);
Eigen::VectorXd scaled_primal_solution_for_residual_tests =
primal_solution_for_residual_tests.cwiseQuotient(col_scaling_vec);
sharded_qp.RescaleQuadraticProgram(col_scaling_vec, row_scaling_vec);
EXPECT_THAT(
ComputeInfeasibilityInformation(
PrimalDualHybridGradientParams(), sharded_qp, col_scaling_vec,
row_scaling_vec, scaled_primal_solution, scaled_dual_solution,
scaled_primal_solution_for_residual_tests,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(expected_infeasibility_info))));
}
TEST(IterationStatsTest, SimpleLpAtOptimum) {
const Eigen::VectorXd primal_solution{{-1.0, 8.0, 1.0, 2.5}};
const Eigen::VectorXd dual_solution{{-2.0, 0.0, 2.375, 2.0 / 3}};
CheckScaledAndUnscaledConvergenceInformation(
TestLp(), primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
ParseTextOrDie<ConvergenceInformation>(R"pb(
primal_objective: -34.0
dual_objective: -34.0
corrected_dual_objective: -34.0
l_inf_primal_residual: 0.0
l2_primal_residual: 0.0
l_inf_componentwise_primal_residual: 0.0
l_inf_dual_residual: 0.0
l2_dual_residual: 0.0
l_inf_componentwise_dual_residual: 0.0
l_inf_primal_variable: 8.0
l2_primal_variable: 8.5
l_inf_dual_variable: 2.375
l2_dual_variable: 3.1756998353818715
)pb"));
}
TEST(IterationStatsTest, SimpleLpWithPrimalResidual) {
// This is the optimal solution, except that x_3 (`primal_solution[3]`) has
// been changed from 2.5 to 3.5, increasing the objective by 1, but causing
// the first constraint to be violated by 2 and the last constraint by 1.
const Eigen::VectorXd primal_solution{{-1.0, 8.0, 1.0, 3.5}};
const Eigen::VectorXd dual_solution{{-2.0, 0.0, 2.375, 2.0 / 3}};
CheckScaledAndUnscaledConvergenceInformation(
TestLp(), primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
ParseTextOrDie<ConvergenceInformation>(R"pb(
primal_objective: -33.0
dual_objective: -34.0
corrected_dual_objective: -34.0
l_inf_primal_residual: 2.0
l2_primal_residual: 2.2360679774997896
l_inf_componentwise_primal_residual: 0.5
l_inf_dual_residual: 0.0
l2_dual_residual: 0.0
l_inf_componentwise_dual_residual: 0.0
l_inf_primal_variable: 8.0
l2_primal_variable: 8.8459030064770662
l_inf_dual_variable: 2.375
l2_dual_variable: 3.1756998353818715
)pb"));
}
TEST(IterationStatsTest, SimpleLpWithDualResidual) {
// This is the optimal solution, except that y_1 (`dual_solution[1]`) has been
// changed from 0 to -1, causing x_0 and x_2 to have primal gradients (dual
// residuals) of 1.0.
const Eigen::VectorXd primal_solution{{-1.0, 8.0, 1.0, 2.5}};
const Eigen::VectorXd dual_solution{{-2.0, -1.0, 2.375, 2.0 / 3}};
CheckScaledAndUnscaledConvergenceInformation(
TestLp(), primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
ParseTextOrDie<ConvergenceInformation>(R"pb(
primal_objective: -34.0
dual_objective: -41.0
corrected_dual_objective: -inf
l_inf_primal_residual: 0.0
l2_primal_residual: 0.0
l_inf_componentwise_primal_residual: 0.0
l_inf_dual_residual: 1.0
l2_dual_residual: 1.4142135623730950
l_inf_componentwise_dual_residual: 0.5
l_inf_primal_variable: 8.0
l2_primal_variable: 8.5
l_inf_dual_variable: 2.375
l2_dual_variable: 3.3294247918288294
)pb"));
}
TEST(IterationStatsTest, SimpleLpWithBothResiduals) {
// This is the optimal solution, except that x_3 (`primal_solution[3]`) has
// been changed from 2.5 to 3.5, increasing the objective by 1, but causing
// the first constraint to be violated by 2 and the last constraint by 1, and
// y_1 (`dual_solution[1]`) has been changed from 0 to -1, causing x_0 and x_2
// to have primal gradients (dual residuals) of 1.0. The primal and dual
// componentwise_residual_offset values are different, to check that the
// correct offset is applied when computing the
// l_inf_componentwise_XXX_residual values.
const Eigen::VectorXd primal_solution{{-1.0, 8.0, 1.0, 3.5}};
const Eigen::VectorXd dual_solution{{-2.0, -1.0, 2.375, 2.0 / 3}};
CheckScaledAndUnscaledConvergenceInformation(
TestLp(), primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/3.0,
/*componentwise_dual_residual_offset=*/1.0,
ParseTextOrDie<ConvergenceInformation>(R"pb(
primal_objective: -33.0
dual_objective: -41.0
corrected_dual_objective: -inf
l_inf_primal_residual: 2.0
l2_primal_residual: 2.2360679774997896
l_inf_componentwise_primal_residual: 0.25
l_inf_dual_residual: 1.0
l2_dual_residual: 1.4142135623730950
l_inf_componentwise_dual_residual: 0.5
l_inf_primal_variable: 8.0
l2_primal_variable: 8.8459030064770662
l_inf_dual_variable: 2.375
l2_dual_variable: 3.3294247918288294
)pb"));
}
TEST(IterationStatsTest, SimpleQpAtOptimum) {
const Eigen::VectorXd primal_solution{{1.0, 0.0}};
const Eigen::VectorXd dual_solution{{-1.0}};
CheckScaledAndUnscaledConvergenceInformation(
TestDiagonalQp1(), primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
ParseTextOrDie<ConvergenceInformation>(R"pb(
primal_objective: 6.0
dual_objective: 6.0
corrected_dual_objective: 6.0
l_inf_primal_residual: 0.0
l2_primal_residual: 0.0
l_inf_componentwise_primal_residual: 0.0
l_inf_dual_residual: 0.0
l2_dual_residual: 0.0
l_inf_componentwise_dual_residual: 0.0
l_inf_primal_variable: 1.0
l2_primal_variable: 1.0
l_inf_dual_variable: 1.0
l2_dual_variable: 1.0
)pb"));
}
TEST(IterationStatsTest, SimpleLpWithGapResidualsAndZeroPrimalSolution) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
const Eigen::VectorXd primal_solution = Eigen::VectorXd::Zero(4);
const Eigen::VectorXd dual_solution{{1.0, 0.0, 0.0, -1.0}};
PrimalDualHybridGradientParams params_true, params_false;
params_true.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
true);
params_false.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
false);
// c is: [5.5, -2, -1, 1]
// -A^T y is: [-2, -1, 0.5, -3]
// c - A^T y is: [3.5, -3.0, -0.5, -2.0].
// When the primal variable is 0.0 and the bound is not 0.0, the bound
// corresponding to c - A^T y is handled as infinite when
// `handle_some_primal_gradients_on_finite_bounds_as_residuals` is true.
// Thus, for the all zero primal solution: when
// `handle_some_primal_gradients_on_finite_bounds_as_residuals` is true, the
// residuals are [3.5, -3.0, -0.5, -2.0] and all bounds are treated as
// infinite. When `handle_some_primal_gradients_on_finite_bounds_as_residuals`
// is false, the residuals are [3.5, -3.0, 0, 0] and the corresponding bound
// terms are [0.0, -2, 6, 3.5].
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_true, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: -3.0
corrected_dual_objective: -inf
l_inf_dual_residual: 3.5
# 5.0497524691810389 = L_2(3.5, -3.0, -0.5, -2.0)
l2_dual_residual: 5.0497524691810389
)pb"))));
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_false, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: -7.0
corrected_dual_objective: -inf
l_inf_dual_residual: 3.5
# 4.6097722286464436 = L_2(3.5, -3.0, 0.0, 0.0)
l2_dual_residual: 4.6097722286464436
)pb"))));
}
TEST(IterationStatsTest, SimpleLpWithGapResidualsAndNonZeroPrimalSolution) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
const Eigen::VectorXd primal_solution{{0.0, 0.0, 4.0, 3.0}};
const Eigen::VectorXd dual_solution{{1.0, 0.0, 0.0, -1.0}};
PrimalDualHybridGradientParams params_true, params_false;
params_true.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
true);
params_false.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
false);
// c is: [5.5, -2, -1, 1]
// -A^T y is: [-2, -1, 0.5, -3]
// c - A^T y is: [3.5, -3.0, -0.5, -2.0].
// When the primal variable is 0.0 and the bound is not 0.0, the bound
// corresponding to c - A^T y is treated as infinite when
// `handle_some_primal_gradients_on_finite_bounds_as_residuals` is true.
// Thus, for primal_solution [0, 0, 4, 3], whether
// `handle_some_primal_gradients_on_finite_bounds_as_residuals` is true or
// not, the residuals are [3.5, -3.0, 0.0, 0.0] and the corresponding bound
// terms are [0.0, -2, 6, 3.5].
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_true, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: -13.0
corrected_dual_objective: -inf
l_inf_dual_residual: 3.5
# 4.6097722286464436 = L_2(3.5, -3.0, 0.0, 0.0)
l2_dual_residual: 4.6097722286464436
)pb"))));
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_false, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: -7.0
corrected_dual_objective: -inf
l_inf_dual_residual: 3.5
# 4.6097722286464436 = L_2(3.5, -3.0, 0.0, 0.0)
l2_dual_residual: 4.6097722286464436
)pb"))));
}
TEST(IterationStatsTest, SimpleQp) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestDiagonalQp1(), num_threads,
num_shards);
const Eigen::VectorXd primal_solution{{1.0, 2.0}};
const Eigen::VectorXd dual_solution{{0.0}};
PrimalDualHybridGradientParams params_true, params_false;
params_true.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
true);
params_false.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(
false);
// Q*x is: [4.0, 2.0]
// c is: [-1, -1]
// A^T y is zero.
// If `handle_some_primal_gradients_on_finite_bounds_as_residuals` is
// true the second primal gradient term is handled as a residual, not a
// reduced cost.
// Other than the reduced cost terms, the dual objective is 5 (objective
// offset) - 4 (1/2 x^T Q x) = 1
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_true, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: 8
corrected_dual_objective: 2
l_inf_dual_residual: 1.0
l2_dual_residual: 1.0
)pb"))));
EXPECT_THAT(ComputeScaledConvergenceInformation(
params_false, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0,
POINT_TYPE_CURRENT_ITERATE),
Partially(Approximately(EqualsProto(R"pb(
dual_objective: 2
corrected_dual_objective: 2
l_inf_dual_residual: 0.0
l2_dual_residual: 0.0
)pb"))));
}
TEST(IterationStatsTest, InfeasibilityInformationWithCertificateLp) {
const Eigen::VectorXd primal_ray{{0.0, 0.0}};
const Eigen::VectorXd dual_ray{{-1.0, -1.0}};
CheckScaledAndUnscaledInfeasibilityStats(
SmallPrimalInfeasibleLp(), primal_ray, dual_ray, primal_ray,
ParseTextOrDie<InfeasibilityInformation>(R"pb(
max_primal_ray_infeasibility: 0
primal_ray_linear_objective: 0
primal_ray_quadratic_norm: 0
max_dual_ray_infeasibility: 0
dual_ray_objective: 1
)pb"));
}
TEST(IterationStatsTest, InfeasibilityInformationWithoutCertificateLp) {
const Eigen::VectorXd primal_ray{{2.0, 1.0}};
const Eigen::VectorXd dual_ray{{-1.0, -3.0}};
CheckScaledAndUnscaledInfeasibilityStats(
SmallPrimalInfeasibleLp(), primal_ray, dual_ray, primal_ray,
ParseTextOrDie<InfeasibilityInformation>(R"pb(
max_primal_ray_infeasibility: 0.5
primal_ray_linear_objective: 1.5
primal_ray_quadratic_norm: 0
max_dual_ray_infeasibility: 0.66666666666666663
dual_ray_objective: 1.6666666666666667
)pb"));
}
TEST(IterationStatsTest, DetectsDualRayHasInfeasibleComponent) {
const Eigen::VectorXd primal_ray{{0.0, 0.0}};
const Eigen::VectorXd dual_ray{{1.0, 1.0}};
// Components with the wrong sign cause the dual ray objective to be -inf.
CheckScaledAndUnscaledInfeasibilityStats(
SmallPrimalInfeasibleLp(), primal_ray, dual_ray, primal_ray,
ParseTextOrDie<InfeasibilityInformation>(R"pb(
max_dual_ray_infeasibility: 0.0
dual_ray_objective: -inf
)pb"));
}
// Regression test for failures of math_opt's
// SimpleLpTest.OptimalAfterInfeasible test.
TEST(IterationStatsTest, HandlesReducedCostsOnDualRayCorrectly) {
// A trivial LP mimicking the one used in math_opt's test:
// min x
// Constraint: 2 <= x
// Variable: 0 <= x <= 1
QuadraticProgram lp(1, 1);
lp.objective_vector = Eigen::VectorXd{{1}};
lp.constraint_lower_bounds = Eigen::VectorXd{{2}};
lp.constraint_upper_bounds =
Eigen::VectorXd{{std::numeric_limits<double>::infinity()}};
lp.variable_lower_bounds = Eigen::VectorXd{{0}};
lp.variable_upper_bounds = Eigen::VectorXd{{1}};
lp.constraint_matrix.coeffRef(0, 0) = 1.0;
lp.constraint_matrix.makeCompressed();
const Eigen::VectorXd primal_solution{{1.0}};
const Eigen::VectorXd primal_ray{{0.0}};
const Eigen::VectorXd dual_ray{{1.0}};
// `dual_ray_objective` = 2 (objective term) - 1 (reduced cost on x) = 1.
CheckScaledAndUnscaledInfeasibilityStats(
lp, primal_ray, dual_ray, primal_solution,
ParseTextOrDie<InfeasibilityInformation>(R"pb(
max_dual_ray_infeasibility: 0.0
dual_ray_objective: 1.0
)pb"));
}
TEST(CorrectedDualTest, SimpleLpWithSuboptimalDual) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
// Set the primal variables that have primal gradients at their bounds, so
// that the primal gradients are reduced costs.
const Eigen::VectorXd primal_solution{{0, 0, 6, 2.5}};
const Eigen::VectorXd dual_solution{{-2, 0, 2.375, 1}};
const ConvergenceInformation stats = ComputeScaledConvergenceInformation(
PrimalDualHybridGradientParams(), sharded_qp, primal_solution,
dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0, POINT_TYPE_CURRENT_ITERATE);
// -36.5 = -14 - 24 - 9.5 - 1 - 3 + 15
EXPECT_DOUBLE_EQ(stats.dual_objective(), -36.5);
EXPECT_DOUBLE_EQ(stats.corrected_dual_objective(), -36.5);
}
// This is similar to `SimpleLpWithSuboptimalDual`, except with
// x_2 = 2. In the dual correction calculation, the corresponding bound is 6, so
// the primal gradient will be treated as a residual of 0.5 instead of a dual
// correction of -3, but in the corrected dual objective it is still treated as
// a dual correction.
TEST(CorrectedDualTest, SimpleLpWithVariableFarFromBoundAsResiduals) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
const Eigen::VectorXd primal_solution{{0, 0, 2, 2.5}};
const Eigen::VectorXd dual_solution{{-2, 0, 2.375, 1}};
PrimalDualHybridGradientParams params;
params.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(true);
const ConvergenceInformation stats = ComputeScaledConvergenceInformation(
params, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0, POINT_TYPE_CURRENT_ITERATE);
// -33.5 = -14 - 24 - 9.5 - 1 + 15
EXPECT_DOUBLE_EQ(stats.dual_objective(), -33.5);
EXPECT_DOUBLE_EQ(stats.corrected_dual_objective(), -36.5);
EXPECT_DOUBLE_EQ(stats.l_inf_dual_residual(), 0.5);
EXPECT_DOUBLE_EQ(stats.l2_dual_residual(), 0.5);
EXPECT_DOUBLE_EQ(stats.l_inf_componentwise_dual_residual(), 0.25);
}
TEST(CorrectedDualTest, SimpleLpWithVariableFarFromBoundAsReducedCosts) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
const Eigen::VectorXd primal_solution{{0, 0, 2, 2.5}};
const Eigen::VectorXd dual_solution{{-2, 0, 2.375, 1}};
PrimalDualHybridGradientParams params;
params.set_handle_some_primal_gradients_on_finite_bounds_as_residuals(false);
const ConvergenceInformation stats = ComputeScaledConvergenceInformation(
params, sharded_qp, primal_solution, dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0, POINT_TYPE_CURRENT_ITERATE);
// -36.5 = -14 - 24 - 9.5 - 1 - 3 + 15
EXPECT_DOUBLE_EQ(stats.dual_objective(), -36.5);
EXPECT_DOUBLE_EQ(stats.corrected_dual_objective(), -36.5);
EXPECT_DOUBLE_EQ(stats.l_inf_dual_residual(), 0.0);
EXPECT_DOUBLE_EQ(stats.l2_dual_residual(), 0.0);
EXPECT_DOUBLE_EQ(stats.l_inf_componentwise_dual_residual(), 0.0);
}
TEST(CorrectedDualObjective, QpSuboptimal) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestDiagonalQp1(), num_threads,
num_shards);
const Eigen::VectorXd primal_solution{{-2.0, 2.0}};
const Eigen::VectorXd dual_solution{{-3}};
const ConvergenceInformation stats = ComputeScaledConvergenceInformation(
PrimalDualHybridGradientParams(), sharded_qp, primal_solution,
dual_solution,
/*componentwise_primal_residual_offset=*/1.0,
/*componentwise_dual_residual_offset=*/1.0, POINT_TYPE_CURRENT_ITERATE);
// primal gradient vector: [-6, 4]
// Constant term: 5
// Quadratic term: -(16+4)/2 = -10
// Dual objective term: -3 * 1
// Primal variables at bounds term: 2*-6 + -2*4 = -20
// -28.0 = 5 - 10 - 3 - 20
EXPECT_DOUBLE_EQ(stats.corrected_dual_objective(), -28.0);
}
TEST(RandomProjectionsTest, OneRandomProjectionsOfZeroVector) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
PointMetadata metadata;
SetRandomProjections(sharded_qp, /*primal_solution=*/Eigen::VectorXd::Zero(4),
/*dual_solution=*/Eigen::VectorXd::Zero(4),
/*random_projection_seeds=*/{1}, metadata);
EXPECT_THAT(metadata.random_primal_projections(), ElementsAre(0.0));
EXPECT_THAT(metadata.random_dual_projections(), ElementsAre(0.0));
}
TEST(RandomProjectionsTest, TwoRandomProjectionsOfVector) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
PointMetadata metadata;
SetRandomProjections(sharded_qp, /*primal_solution=*/Eigen::VectorXd::Ones(4),
/*dual_solution=*/Eigen::VectorXd::Zero(4),
/*random_projection_seeds=*/{1, 2}, metadata);
EXPECT_THAT(metadata.random_primal_projections(), SizeIs(2));
EXPECT_THAT(metadata.random_dual_projections(), SizeIs(2));
// The primal solution has norm 2; the random projection should only reduce
// the norm. Obtaining 0.0 is a probability-zero event.
EXPECT_THAT(metadata.random_primal_projections(),
Each(AllOf(Ge(-2.0), Le(2.0), Ne(0.0))));
EXPECT_THAT(metadata.random_dual_projections(), Each(Eq(0.0)));
}
TEST(ReducedCostsTest, SimpleLp) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestLp(), num_threads, num_shards);
// Use a primal solution at the relevant bounds, to ensure handling as
// reduced costs.
const Eigen::VectorXd primal_solution{{0.0, -2.0, 6.0, 3.5}};
const Eigen::VectorXd dual_solution{{1.0, 0.0, 0.0, -2.0}};
// c is: [5.5, -2, -1, 1]
// -A^T y is: [-2, -1, 2, -4]
// c - A^T y is: [3.5, -3.0, 1.0, -3.0].
EXPECT_THAT(ReducedCosts(PrimalDualHybridGradientParams(), sharded_qp,
primal_solution, dual_solution),
ElementsAre(3.5, -3.0, 1.0, -3.0));
EXPECT_THAT(ReducedCosts(PrimalDualHybridGradientParams(), sharded_qp,
primal_solution, dual_solution,
/*use_zero_primal_objective=*/true),
ElementsAre(-2.0, -1.0, 2.0, -4.0));
}
TEST(ReducedCostsTest, SimpleQp) {
const int num_threads = 2;
const int num_shards = 10;
ShardedQuadraticProgram sharded_qp(TestDiagonalQp1(), num_threads,
num_shards);
const Eigen::VectorXd primal_solution{{1.0, 2.0}};
const Eigen::VectorXd dual_solution{{0.0}};
// Q*x is: [4.0, 2.0]
// c is: [-1, -1]
// A^T y is zero.
// If `handle_some_primal_gradients_on_finite_bounds_as_residuals` is
// true the second primal gradient term is handled as a residual, not a
// reduced cost.
EXPECT_THAT(ReducedCosts(PrimalDualHybridGradientParams(), sharded_qp,
primal_solution, dual_solution),
ElementsAre(3.0, 1.0));
EXPECT_THAT(ReducedCosts(PrimalDualHybridGradientParams(), sharded_qp,
primal_solution, dual_solution,
/*use_zero_primal_objective=*/true),
ElementsAre(0.0, 0.0));
}
TEST(GetConvergenceInformation, GetsCorrectEntry) {
const auto test_stats = ParseTextOrDie<IterationStats>(R"pb(
convergence_information {
candidate_type: POINT_TYPE_CURRENT_ITERATE
primal_objective: 1.0
}
convergence_information {
candidate_type: POINT_TYPE_AVERAGE_ITERATE
primal_objective: 2.0
}
)pb");
const auto average_info =
GetConvergenceInformation(test_stats, POINT_TYPE_AVERAGE_ITERATE);
ASSERT_TRUE(average_info.has_value());
EXPECT_EQ(average_info->candidate_type(), POINT_TYPE_AVERAGE_ITERATE);
EXPECT_EQ(average_info->primal_objective(), 2.0);
const auto current_info =
GetConvergenceInformation(test_stats, POINT_TYPE_CURRENT_ITERATE);
ASSERT_TRUE(current_info.has_value());
EXPECT_EQ(current_info->candidate_type(), POINT_TYPE_CURRENT_ITERATE);
EXPECT_EQ(current_info->primal_objective(), 1.0);
EXPECT_THAT(
GetConvergenceInformation(test_stats, POINT_TYPE_ITERATE_DIFFERENCE),
Eq(std::nullopt));
}
TEST(GetInfeasibilityInformation, GetsCorrectEntry) {
const auto test_stats = ParseTextOrDie<IterationStats>(R"pb(
infeasibility_information {
candidate_type: POINT_TYPE_CURRENT_ITERATE
primal_ray_linear_objective: 1.0
}
infeasibility_information {
candidate_type: POINT_TYPE_AVERAGE_ITERATE
primal_ray_linear_objective: 2.0
}
)pb");
const auto average_info =
GetInfeasibilityInformation(test_stats, POINT_TYPE_AVERAGE_ITERATE);
ASSERT_TRUE(average_info.has_value());
EXPECT_EQ(average_info->candidate_type(), POINT_TYPE_AVERAGE_ITERATE);
EXPECT_EQ(average_info->primal_ray_linear_objective(), 2.0);
const auto current_info =
GetInfeasibilityInformation(test_stats, POINT_TYPE_CURRENT_ITERATE);
ASSERT_TRUE(current_info.has_value());
EXPECT_EQ(current_info->candidate_type(), POINT_TYPE_CURRENT_ITERATE);
EXPECT_EQ(current_info->primal_ray_linear_objective(), 1.0);
EXPECT_THAT(
GetInfeasibilityInformation(test_stats, POINT_TYPE_ITERATE_DIFFERENCE),
Eq(std::nullopt));
}
TEST(GetPointMetadata, GetsCorrectEntry) {
const auto test_stats = ParseTextOrDie<IterationStats>(R"pb(
point_metadata {
point_type: POINT_TYPE_CURRENT_ITERATE
active_primal_variable_count: 1
}
point_metadata {
point_type: POINT_TYPE_AVERAGE_ITERATE
active_primal_variable_count: 2
}
)pb");
const auto average_info =
GetPointMetadata(test_stats, POINT_TYPE_AVERAGE_ITERATE);
ASSERT_TRUE(average_info.has_value());
EXPECT_EQ(average_info->point_type(), POINT_TYPE_AVERAGE_ITERATE);
EXPECT_EQ(average_info->active_primal_variable_count(), 2);
const auto current_info =
GetPointMetadata(test_stats, POINT_TYPE_CURRENT_ITERATE);
ASSERT_TRUE(current_info.has_value());
EXPECT_EQ(current_info->point_type(), POINT_TYPE_CURRENT_ITERATE);
EXPECT_EQ(current_info->active_primal_variable_count(), 1);
EXPECT_THAT(GetPointMetadata(test_stats, POINT_TYPE_ITERATE_DIFFERENCE),
Eq(std::nullopt));
}
} // namespace
} // namespace operations_research::pdlp