 |
Google OR-Tools v9.9
a fast and portable software suite for combinatorial optimization
|
|
Google OR-Tools v9.9
a fast and portable software suite for combinatorial optimization
|
Static Public Member Functions | |
| static final com.google.protobuf.Descriptors.Descriptor | getDescriptor () |
Protected Member Functions | |
| com.google.protobuf.GeneratedMessageV3.FieldAccessorTable | internalGetFieldAccessorTable () |
Parameters for PrimalDualHybridGradient() in primal_dual_hybrid_gradient.h. While the defaults are generally good, it is usually worthwhile to perform a parameter sweep to find good settings for a particular family of problems. The following parameters should be considered for tuning: - restart_strategy (jointly with major_iteration_frequency) - primal_weight_update_smoothing (jointly with initial_primal_weight) - presolve_options.use_glop - l_inf_ruiz_iterations - l2_norm_rescaling In addition, tune num_threads to speed up the solve.
Protobuf type operations_research.pdlp.PrimalDualHybridGradientParams
Definition at line 9447 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.addAllRandomProjectionSeeds | ( | java.lang.Iterable<? extends java.lang.Integer > | values | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
| values | The randomProjectionSeeds to add. |
Definition at line 11756 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.addRandomProjectionSeeds | ( | int | value | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
| value | The randomProjectionSeeds to add. |
Definition at line 11734 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.addRepeatedField | ( | com.google.protobuf.Descriptors.FieldDescriptor | field, |
| java.lang.Object | value ) |
Definition at line 9705 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.build | ( | ) |
Definition at line 9544 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.buildPartial | ( | ) |
Definition at line 9553 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clear | ( | ) |
Definition at line 9484 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearAdaptiveLinesearchParameters | ( | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Definition at line 11411 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearDiagonalQpTrustRegionSolverTolerance | ( | ) |
The solve tolerance of the experimental trust region solver for diagonal QPs, controlling the accuracy of binary search over a one-dimensional scaling parameter. Smaller values imply smaller relative error of the final solution vector. TODO(user): Find an expression for the final relative error.
optional double diagonal_qp_trust_region_solver_tolerance = 24 [default = 1e-08];
Definition at line 12066 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearField | ( | com.google.protobuf.Descriptors.FieldDescriptor | field | ) |
Definition at line 9689 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearHandleSomePrimalGradientsOnFiniteBoundsAsResiduals | ( | ) |
See https://developers.google.com/optimization/lp/pdlp_math#treating_some_variable_bounds_as_infinite for a description of this flag.
optional bool handle_some_primal_gradients_on_finite_bounds_as_residuals = 29 [default = true];
Definition at line 11918 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearInfiniteConstraintBoundThreshold | ( | ) |
Constraint bounds with absolute value at least this threshold are replaced with infinities. NOTE: This primarily affects the relative convergence criteria. A smaller value makes the relative convergence criteria stronger. It also affects the problem statistics LOG()ed at the start of the run, and the default initial primal weight, since that is based on the norm of the bounds.
optional double infinite_constraint_bound_threshold = 22 [default = inf];
Definition at line 11854 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearInitialPrimalWeight | ( | ) |
The initial value of the primal weight (i.e., the ratio of primal and dual
step sizes). The primal weight remains fixed throughout the solve if
primal_weight_update_smoothing = 0.0. If unset, the default is the ratio of
the norm of the objective vector to the L2 norm of the combined constraint
bounds vector (as defined above). If this ratio is not finite and positive,
then the default is 1.0 instead. For tuning, try powers of 10, for example,
from 10^{-6} to 10^6.
optional double initial_primal_weight = 8;
Definition at line 10890 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearInitialStepSizeScaling | ( | ) |
Scaling factor applied to the initial step size (all step sizes if linesearch_rule == CONSTANT_STEP_SIZE_RULE).
optional double initial_step_size_scaling = 25 [default = 1];
Definition at line 11631 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearL2NormRescaling | ( | ) |
If true, applies L_2 norm rescaling after the Ruiz rescaling. Heuristically this has been found to help convergence.
optional bool l2_norm_rescaling = 10 [default = true];
Definition at line 11135 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearLinesearchRule | ( | ) |
Linesearch rule applied at each major iteration.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.LinesearchRule linesearch_rule = 12 [default = ADAPTIVE_LINESEARCH_RULE];
Definition at line 11329 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearLInfRuizIterations | ( | ) |
Number of L_infinity Ruiz rescaling iterations to apply to the constraint matrix. Zero disables this rescaling pass. Recommended values to try when tuning are 0, 5, and 10.
optional int32 l_inf_ruiz_iterations = 9 [default = 5];
Definition at line 11075 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearLogIntervalSeconds | ( | ) |
Time between iteration-level statistics logging (if `verbosity_level > 1`). Since iteration-level statistics are only generated when performing termination checks, logs will be generated from next termination check after `log_interval_seconds` have elapsed. Should be >= 0.0. 0.0 (the default) means log statistics at every termination check.
optional double log_interval_seconds = 31 [default = 0];
Definition at line 10520 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearMajorIterationFrequency | ( | ) |
The frequency at which extra work is performed to make major algorithmic decisions, e.g., performing restarts and updating the primal weight. Major iterations also trigger a termination check. For best performance using the NO_RESTARTS or EVERY_MAJOR_ITERATION rule, one should perform a log-scale grid search over this parameter, for example, over powers of two. ADAPTIVE_HEURISTIC is mostly insensitive to this value.
optional int32 major_iteration_frequency = 4 [default = 64];
Definition at line 10596 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearMalitskyPockParameters | ( | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Definition at line 11532 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearNecessaryReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC only: A relative reduction in the potential function by this amount triggers a restart if, additionally, the quality of the iterates appears to be getting worse. The value must be in the interval [sufficient_reduction_for_restart, 1). Smaller values make restarts less frequent, and larger values make them more frequent.
optional double necessary_reduction_for_restart = 17 [default = 0.9];
Definition at line 11271 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearNumShards | ( | ) |
For more efficient parallel computation, the matrices and vectors are divided (virtually) into num_shards shards. Results are computed independently for each shard and then combined. As a consequence, the order of computation, and hence floating point roundoff, depends on the number of shards so reproducible results require using the same value for num_shards. However, for efficiency num_shards should a be at least num_threads, and preferably at least 4*num_threads to allow better load balancing. If num_shards is positive, the computation will use that many shards. Otherwise a default that depends on num_threads will be used.
optional int32 num_shards = 27 [default = 0];
Definition at line 10284 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearNumThreads | ( | ) |
The number of threads to use. Must be positive. Try various values of num_threads, up to the number of physical cores. Performance may not be monotonically increasing with the number of threads because of memory bandwidth limitations.
optional int32 num_threads = 2 [default = 1];
Definition at line 10196 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearOneof | ( | com.google.protobuf.Descriptors.OneofDescriptor | oneof | ) |
Definition at line 9694 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearPresolveOptions | ( | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Definition at line 10972 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearPrimalWeightUpdateSmoothing | ( | ) |
This parameter controls exponential smoothing of log(primal_weight) when a primal weight update occurs (i.e., when the ratio of primal and dual step sizes is adjusted). At 0.0, the primal weight will be frozen at its initial value and there will be no dynamic updates in the algorithm. At 1.0, there is no smoothing in the updates. The default of 0.5 generally performs well, but has been observed on occasion to trigger unstable swings in the primal weight. We recommend also trying 0.0 (disabling primal weight updates), in which case you must also tune initial_primal_weight.
optional double primal_weight_update_smoothing = 7 [default = 0.5];
Definition at line 10810 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearRandomProjectionSeeds | ( | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
Definition at line 11778 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearRecordIterationStats | ( | ) |
If true, the iteration_stats field of the SolveLog output will be populated at every iteration. Note that we only compute solution statistics at termination checks. Setting this parameter to true may substantially increase the size of the output.
optional bool record_iteration_stats = 3;
Definition at line 10352 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearRestartStrategy | ( | ) |
NO_RESTARTS and EVERY_MAJOR_ITERATION occasionally outperform the default. If using a strategy other than ADAPTIVE_HEURISTIC, you must also tune major_iteration_frequency.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.RestartStrategy restart_strategy = 6 [default = ADAPTIVE_HEURISTIC];
Definition at line 10726 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearSufficientReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC and ADAPTIVE_DISTANCE_BASED only: A relative reduction in the potential function by this amount always triggers a restart. Must be between 0.0 and 1.0.
optional double sufficient_reduction_for_restart = 11 [default = 0.1];
Definition at line 11199 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearTerminationCheckFrequency | ( | ) |
The frequency (based on a counter reset every major iteration) to check for termination (involves extra work) and log iteration stats. Termination checks do not affect algorithmic progress unless termination is triggered.
optional int32 termination_check_frequency = 5 [default = 64];
Definition at line 10660 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearTerminationCriteria | ( | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Definition at line 10089 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearUseDiagonalQpTrustRegionSolver | ( | ) |
When solving QPs with diagonal objective matrices, this option can be turned on to enable an experimental solver that avoids linearization of the quadratic term. The `diagonal_qp_solver_accuracy` parameter controls the solve accuracy. TODO(user): Turn this option on by default for quadratic programs after numerical evaluation.
optional bool use_diagonal_qp_trust_region_solver = 23 [default = false];
Definition at line 11994 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearUseFeasibilityPolishing | ( | ) |
If true, periodically runs feasibility polishing, which attempts to move from latest average iterate to one that is closer to feasibility (i.e., has smaller primal and dual residuals) while probably increasing the objective gap. This is useful primarily when the feasibility tolerances are fairly tight and the objective gap tolerance is somewhat looser. Note that this does not change the termination criteria, but rather can help achieve the termination criteria more quickly when the objective gap is not as important as feasibility. `use_feasibility_polishing` cannot be used with glop presolve, and requires `handle_some_primal_gradients_on_finite_bounds_as_residuals == false`. `use_feasibility_polishing` can only be used with linear programs. Feasibility polishing runs two separate phases, primal feasibility and dual feasibility. The primal feasibility phase runs PDHG on the primal feasibility problem (obtained by changing the objective vector to all zeros), using the average primal iterate and zero dual (which is optimal for the primal feasibility problem) as the initial solution. The dual feasibility phase runs PDHG on the dual feasibility problem (obtained by changing all finite variable and constraint bounds to zero), using the average dual iterate and zero primal (which is optimal for the dual feasibility problem) as the initial solution. The primal solution from the primal feasibility phase and dual solution from the dual feasibility phase are then combined (forming a solution of type `POINT_TYPE_FEASIBILITY_POLISHING_SOLUTION`) and checked against the termination criteria.
optional bool use_feasibility_polishing = 30 [default = false];
Definition at line 12222 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clearVerbosityLevel | ( | ) |
The verbosity of logging. 0: No informational logging. (Errors are logged.) 1: Summary statistics only. No iteration-level details. 2: A table of iteration-level statistics is logged. (See ToShortString() in primal_dual_hybrid_gradient.cc). 3: A more detailed table of iteration-level statistics is logged. (See ToString() in primal_dual_hybrid_gradient.cc). 4: For iteration-level details, prints the statistics of both the average (prefixed with A) and the current iterate (prefixed with C). Also prints internal algorithmic state and details. Logging at levels 2-4 also includes messages from level 1.
optional int32 verbosity_level = 26 [default = 0];
Definition at line 10448 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.clone | ( | ) |
Definition at line 9679 of file Solvers.java.
| operations_research.pdlp.Solvers.AdaptiveLinesearchParams operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getAdaptiveLinesearchParameters | ( | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11350 of file Solvers.java.
| operations_research.pdlp.Solvers.AdaptiveLinesearchParams.Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getAdaptiveLinesearchParametersBuilder | ( | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Definition at line 11424 of file Solvers.java.
| operations_research.pdlp.Solvers.AdaptiveLinesearchParamsOrBuilder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getAdaptiveLinesearchParametersOrBuilder | ( | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11432 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getDefaultInstanceForType | ( | ) |
Definition at line 9539 of file Solvers.java.
|
static |
Definition at line 9452 of file Solvers.java.
| com.google.protobuf.Descriptors.Descriptor operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getDescriptorForType | ( | ) |
Definition at line 9534 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getDiagonalQpTrustRegionSolverTolerance | ( | ) |
The solve tolerance of the experimental trust region solver for diagonal QPs, controlling the accuracy of binary search over a one-dimensional scaling parameter. Smaller values imply smaller relative error of the final solution vector. TODO(user): Find an expression for the final relative error.
optional double diagonal_qp_trust_region_solver_tolerance = 24 [default = 1e-08];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 12031 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getHandleSomePrimalGradientsOnFiniteBoundsAsResiduals | ( | ) |
See https://developers.google.com/optimization/lp/pdlp_math#treating_some_variable_bounds_as_infinite for a description of this flag.
optional bool handle_some_primal_gradients_on_finite_bounds_as_residuals = 29 [default = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11887 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getInfiniteConstraintBoundThreshold | ( | ) |
Constraint bounds with absolute value at least this threshold are replaced with infinities. NOTE: This primarily affects the relative convergence criteria. A smaller value makes the relative convergence criteria stronger. It also affects the problem statistics LOG()ed at the start of the run, and the default initial primal weight, since that is based on the norm of the bounds.
optional double infinite_constraint_bound_threshold = 22 [default = inf];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11817 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getInitialPrimalWeight | ( | ) |
The initial value of the primal weight (i.e., the ratio of primal and dual
step sizes). The primal weight remains fixed throughout the solve if
primal_weight_update_smoothing = 0.0. If unset, the default is the ratio of
the norm of the objective vector to the L2 norm of the combined constraint
bounds vector (as defined above). If this ratio is not finite and positive,
then the default is 1.0 instead. For tuning, try powers of 10, for example,
from 10^{-6} to 10^6.
optional double initial_primal_weight = 8;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10851 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getInitialStepSizeScaling | ( | ) |
Scaling factor applied to the initial step size (all step sizes if linesearch_rule == CONSTANT_STEP_SIZE_RULE).
optional double initial_step_size_scaling = 25 [default = 1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11602 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getL2NormRescaling | ( | ) |
If true, applies L_2 norm rescaling after the Ruiz rescaling. Heuristically this has been found to help convergence.
optional bool l2_norm_rescaling = 10 [default = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11106 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.LinesearchRule operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getLinesearchRule | ( | ) |
Linesearch rule applied at each major iteration.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.LinesearchRule linesearch_rule = 12 [default = ADAPTIVE_LINESEARCH_RULE];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11299 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getLInfRuizIterations | ( | ) |
Number of L_infinity Ruiz rescaling iterations to apply to the constraint matrix. Zero disables this rescaling pass. Recommended values to try when tuning are 0, 5, and 10.
optional int32 l_inf_ruiz_iterations = 9 [default = 5];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11044 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getLogIntervalSeconds | ( | ) |
Time between iteration-level statistics logging (if `verbosity_level > 1`). Since iteration-level statistics are only generated when performing termination checks, logs will be generated from next termination check after `log_interval_seconds` have elapsed. Should be >= 0.0. 0.0 (the default) means log statistics at every termination check.
optional double log_interval_seconds = 31 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10485 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getMajorIterationFrequency | ( | ) |
The frequency at which extra work is performed to make major algorithmic decisions, e.g., performing restarts and updating the primal weight. Major iterations also trigger a termination check. For best performance using the NO_RESTARTS or EVERY_MAJOR_ITERATION rule, one should perform a log-scale grid search over this parameter, for example, over powers of two. ADAPTIVE_HEURISTIC is mostly insensitive to this value.
optional int32 major_iteration_frequency = 4 [default = 64];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10559 of file Solvers.java.
| operations_research.pdlp.Solvers.MalitskyPockParams operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getMalitskyPockParameters | ( | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11471 of file Solvers.java.
| operations_research.pdlp.Solvers.MalitskyPockParams.Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getMalitskyPockParametersBuilder | ( | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Definition at line 11545 of file Solvers.java.
| operations_research.pdlp.Solvers.MalitskyPockParamsOrBuilder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getMalitskyPockParametersOrBuilder | ( | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11553 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getNecessaryReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC only: A relative reduction in the potential function by this amount triggers a restart if, additionally, the quality of the iterates appears to be getting worse. The value must be in the interval [sufficient_reduction_for_restart, 1). Smaller values make restarts less frequent, and larger values make them more frequent.
optional double necessary_reduction_for_restart = 17 [default = 0.9];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11236 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getNumShards | ( | ) |
For more efficient parallel computation, the matrices and vectors are divided (virtually) into num_shards shards. Results are computed independently for each shard and then combined. As a consequence, the order of computation, and hence floating point roundoff, depends on the number of shards so reproducible results require using the same value for num_shards. However, for efficiency num_shards should a be at least num_threads, and preferably at least 4*num_threads to allow better load balancing. If num_shards is positive, the computation will use that many shards. Otherwise a default that depends on num_threads will be used.
optional int32 num_shards = 27 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10241 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getNumThreads | ( | ) |
The number of threads to use. Must be positive. Try various values of num_threads, up to the number of physical cores. Performance may not be monotonically increasing with the number of threads because of memory bandwidth limitations.
optional int32 num_threads = 2 [default = 1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10163 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptions operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getPresolveOptions | ( | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10911 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptions.Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getPresolveOptionsBuilder | ( | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Definition at line 10985 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptionsOrBuilder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getPresolveOptionsOrBuilder | ( | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10993 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getPrimalWeightUpdateSmoothing | ( | ) |
This parameter controls exponential smoothing of log(primal_weight) when a primal weight update occurs (i.e., when the ratio of primal and dual step sizes is adjusted). At 0.0, the primal weight will be frozen at its initial value and there will be no dynamic updates in the algorithm. At 1.0, there is no smoothing in the updates. The default of 0.5 generally performs well, but has been observed on occasion to trigger unstable swings in the primal weight. We recommend also trying 0.0 (disabling primal weight updates), in which case you must also tune initial_primal_weight.
optional double primal_weight_update_smoothing = 7 [default = 0.5];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10769 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getRandomProjectionSeeds | ( | int | index | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
| index | The index of the element to return. |
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11693 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getRandomProjectionSeedsCount | ( | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11676 of file Solvers.java.
| java.util.List< java.lang.Integer > operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getRandomProjectionSeedsList | ( | ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11659 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getRecordIterationStats | ( | ) |
If true, the iteration_stats field of the SolveLog output will be populated at every iteration. Note that we only compute solution statistics at termination checks. Setting this parameter to true may substantially increase the size of the output.
optional bool record_iteration_stats = 3;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10319 of file Solvers.java.
| operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.RestartStrategy operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getRestartStrategy | ( | ) |
NO_RESTARTS and EVERY_MAJOR_ITERATION occasionally outperform the default. If using a strategy other than ADAPTIVE_HEURISTIC, you must also tune major_iteration_frequency.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.RestartStrategy restart_strategy = 6 [default = ADAPTIVE_HEURISTIC];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10692 of file Solvers.java.
| double operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getSufficientReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC and ADAPTIVE_DISTANCE_BASED only: A relative reduction in the potential function by this amount always triggers a restart. Must be between 0.0 and 1.0.
optional double sufficient_reduction_for_restart = 11 [default = 0.1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11168 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getTerminationCheckFrequency | ( | ) |
The frequency (based on a counter reset every major iteration) to check for termination (involves extra work) and log iteration stats. Termination checks do not affect algorithmic progress unless termination is triggered.
optional int32 termination_check_frequency = 5 [default = 64];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10629 of file Solvers.java.
| operations_research.pdlp.Solvers.TerminationCriteria operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getTerminationCriteria | ( | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10028 of file Solvers.java.
| operations_research.pdlp.Solvers.TerminationCriteria.Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getTerminationCriteriaBuilder | ( | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Definition at line 10102 of file Solvers.java.
| operations_research.pdlp.Solvers.TerminationCriteriaOrBuilder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getTerminationCriteriaOrBuilder | ( | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10110 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getUseDiagonalQpTrustRegionSolver | ( | ) |
When solving QPs with diagonal objective matrices, this option can be turned on to enable an experimental solver that avoids linearization of the quadratic term. The `diagonal_qp_solver_accuracy` parameter controls the solve accuracy. TODO(user): Turn this option on by default for quadratic programs after numerical evaluation.
optional bool use_diagonal_qp_trust_region_solver = 23 [default = false];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11957 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getUseFeasibilityPolishing | ( | ) |
If true, periodically runs feasibility polishing, which attempts to move from latest average iterate to one that is closer to feasibility (i.e., has smaller primal and dual residuals) while probably increasing the objective gap. This is useful primarily when the feasibility tolerances are fairly tight and the objective gap tolerance is somewhat looser. Note that this does not change the termination criteria, but rather can help achieve the termination criteria more quickly when the objective gap is not as important as feasibility. `use_feasibility_polishing` cannot be used with glop presolve, and requires `handle_some_primal_gradients_on_finite_bounds_as_residuals == false`. `use_feasibility_polishing` can only be used with linear programs. Feasibility polishing runs two separate phases, primal feasibility and dual feasibility. The primal feasibility phase runs PDHG on the primal feasibility problem (obtained by changing the objective vector to all zeros), using the average primal iterate and zero dual (which is optimal for the primal feasibility problem) as the initial solution. The dual feasibility phase runs PDHG on the dual feasibility problem (obtained by changing all finite variable and constraint bounds to zero), using the average dual iterate and zero primal (which is optimal for the dual feasibility problem) as the initial solution. The primal solution from the primal feasibility phase and dual solution from the dual feasibility phase are then combined (forming a solution of type `POINT_TYPE_FEASIBILITY_POLISHING_SOLUTION`) and checked against the termination criteria.
optional bool use_feasibility_polishing = 30 [default = false];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 12145 of file Solvers.java.
| int operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.getVerbosityLevel | ( | ) |
The verbosity of logging. 0: No informational logging. (Errors are logged.) 1: Summary statistics only. No iteration-level details. 2: A table of iteration-level statistics is logged. (See ToShortString() in primal_dual_hybrid_gradient.cc). 3: A more detailed table of iteration-level statistics is logged. (See ToString() in primal_dual_hybrid_gradient.cc). 4: For iteration-level details, prints the statistics of both the average (prefixed with A) and the current iterate (prefixed with C). Also prints internal algorithmic state and details. Logging at levels 2-4 also includes messages from level 1.
optional int32 verbosity_level = 26 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10401 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasAdaptiveLinesearchParameters | ( | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11343 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasDiagonalQpTrustRegionSolverTolerance | ( | ) |
The solve tolerance of the experimental trust region solver for diagonal QPs, controlling the accuracy of binary search over a one-dimensional scaling parameter. Smaller values imply smaller relative error of the final solution vector. TODO(user): Find an expression for the final relative error.
optional double diagonal_qp_trust_region_solver_tolerance = 24 [default = 1e-08];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 12015 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasHandleSomePrimalGradientsOnFiniteBoundsAsResiduals | ( | ) |
See https://developers.google.com/optimization/lp/pdlp_math#treating_some_variable_bounds_as_infinite for a description of this flag.
optional bool handle_some_primal_gradients_on_finite_bounds_as_residuals = 29 [default = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11873 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasInfiniteConstraintBoundThreshold | ( | ) |
Constraint bounds with absolute value at least this threshold are replaced with infinities. NOTE: This primarily affects the relative convergence criteria. A smaller value makes the relative convergence criteria stronger. It also affects the problem statistics LOG()ed at the start of the run, and the default initial primal weight, since that is based on the norm of the bounds.
optional double infinite_constraint_bound_threshold = 22 [default = inf];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11800 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasInitialPrimalWeight | ( | ) |
The initial value of the primal weight (i.e., the ratio of primal and dual
step sizes). The primal weight remains fixed throughout the solve if
primal_weight_update_smoothing = 0.0. If unset, the default is the ratio of
the norm of the objective vector to the L2 norm of the combined constraint
bounds vector (as defined above). If this ratio is not finite and positive,
then the default is 1.0 instead. For tuning, try powers of 10, for example,
from 10^{-6} to 10^6.
optional double initial_primal_weight = 8;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10833 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasInitialStepSizeScaling | ( | ) |
Scaling factor applied to the initial step size (all step sizes if linesearch_rule == CONSTANT_STEP_SIZE_RULE).
optional double initial_step_size_scaling = 25 [default = 1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11589 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasL2NormRescaling | ( | ) |
If true, applies L_2 norm rescaling after the Ruiz rescaling. Heuristically this has been found to help convergence.
optional bool l2_norm_rescaling = 10 [default = true];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11093 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasLinesearchRule | ( | ) |
Linesearch rule applied at each major iteration.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.LinesearchRule linesearch_rule = 12 [default = ADAPTIVE_LINESEARCH_RULE];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11287 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasLInfRuizIterations | ( | ) |
Number of L_infinity Ruiz rescaling iterations to apply to the constraint matrix. Zero disables this rescaling pass. Recommended values to try when tuning are 0, 5, and 10.
optional int32 l_inf_ruiz_iterations = 9 [default = 5];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11030 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasLogIntervalSeconds | ( | ) |
Time between iteration-level statistics logging (if `verbosity_level > 1`). Since iteration-level statistics are only generated when performing termination checks, logs will be generated from next termination check after `log_interval_seconds` have elapsed. Should be >= 0.0. 0.0 (the default) means log statistics at every termination check.
optional double log_interval_seconds = 31 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10469 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasMajorIterationFrequency | ( | ) |
The frequency at which extra work is performed to make major algorithmic decisions, e.g., performing restarts and updating the primal weight. Major iterations also trigger a termination check. For best performance using the NO_RESTARTS or EVERY_MAJOR_ITERATION rule, one should perform a log-scale grid search over this parameter, for example, over powers of two. ADAPTIVE_HEURISTIC is mostly insensitive to this value.
optional int32 major_iteration_frequency = 4 [default = 64];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10542 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasMalitskyPockParameters | ( | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11464 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasNecessaryReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC only: A relative reduction in the potential function by this amount triggers a restart if, additionally, the quality of the iterates appears to be getting worse. The value must be in the interval [sufficient_reduction_for_restart, 1). Smaller values make restarts less frequent, and larger values make them more frequent.
optional double necessary_reduction_for_restart = 17 [default = 0.9];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11220 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasNumShards | ( | ) |
For more efficient parallel computation, the matrices and vectors are divided (virtually) into num_shards shards. Results are computed independently for each shard and then combined. As a consequence, the order of computation, and hence floating point roundoff, depends on the number of shards so reproducible results require using the same value for num_shards. However, for efficiency num_shards should a be at least num_threads, and preferably at least 4*num_threads to allow better load balancing. If num_shards is positive, the computation will use that many shards. Otherwise a default that depends on num_threads will be used.
optional int32 num_shards = 27 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10221 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasNumThreads | ( | ) |
The number of threads to use. Must be positive. Try various values of num_threads, up to the number of physical cores. Performance may not be monotonically increasing with the number of threads because of memory bandwidth limitations.
optional int32 num_threads = 2 [default = 1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10148 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasPresolveOptions | ( | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10904 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasPrimalWeightUpdateSmoothing | ( | ) |
This parameter controls exponential smoothing of log(primal_weight) when a primal weight update occurs (i.e., when the ratio of primal and dual step sizes is adjusted). At 0.0, the primal weight will be frozen at its initial value and there will be no dynamic updates in the algorithm. At 1.0, there is no smoothing in the updates. The default of 0.5 generally performs well, but has been observed on occasion to trigger unstable swings in the primal weight. We recommend also trying 0.0 (disabling primal weight updates), in which case you must also tune initial_primal_weight.
optional double primal_weight_update_smoothing = 7 [default = 0.5];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10750 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasRecordIterationStats | ( | ) |
If true, the iteration_stats field of the SolveLog output will be populated at every iteration. Note that we only compute solution statistics at termination checks. Setting this parameter to true may substantially increase the size of the output.
optional bool record_iteration_stats = 3;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10304 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasRestartStrategy | ( | ) |
NO_RESTARTS and EVERY_MAJOR_ITERATION occasionally outperform the default. If using a strategy other than ADAPTIVE_HEURISTIC, you must also tune major_iteration_frequency.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.RestartStrategy restart_strategy = 6 [default = ADAPTIVE_HEURISTIC];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10678 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasSufficientReductionForRestart | ( | ) |
For ADAPTIVE_HEURISTIC and ADAPTIVE_DISTANCE_BASED only: A relative reduction in the potential function by this amount always triggers a restart. Must be between 0.0 and 1.0.
optional double sufficient_reduction_for_restart = 11 [default = 0.1];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11154 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasTerminationCheckFrequency | ( | ) |
The frequency (based on a counter reset every major iteration) to check for termination (involves extra work) and log iteration stats. Termination checks do not affect algorithmic progress unless termination is triggered.
optional int32 termination_check_frequency = 5 [default = 64];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10615 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasTerminationCriteria | ( | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10021 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasUseDiagonalQpTrustRegionSolver | ( | ) |
When solving QPs with diagonal objective matrices, this option can be turned on to enable an experimental solver that avoids linearization of the quadratic term. The `diagonal_qp_solver_accuracy` parameter controls the solve accuracy. TODO(user): Turn this option on by default for quadratic programs after numerical evaluation.
optional bool use_diagonal_qp_trust_region_solver = 23 [default = false];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 11940 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasUseFeasibilityPolishing | ( | ) |
If true, periodically runs feasibility polishing, which attempts to move from latest average iterate to one that is closer to feasibility (i.e., has smaller primal and dual residuals) while probably increasing the objective gap. This is useful primarily when the feasibility tolerances are fairly tight and the objective gap tolerance is somewhat looser. Note that this does not change the termination criteria, but rather can help achieve the termination criteria more quickly when the objective gap is not as important as feasibility. `use_feasibility_polishing` cannot be used with glop presolve, and requires `handle_some_primal_gradients_on_finite_bounds_as_residuals == false`. `use_feasibility_polishing` can only be used with linear programs. Feasibility polishing runs two separate phases, primal feasibility and dual feasibility. The primal feasibility phase runs PDHG on the primal feasibility problem (obtained by changing the objective vector to all zeros), using the average primal iterate and zero dual (which is optimal for the primal feasibility problem) as the initial solution. The dual feasibility phase runs PDHG on the dual feasibility problem (obtained by changing all finite variable and constraint bounds to zero), using the average dual iterate and zero primal (which is optimal for the dual feasibility problem) as the initial solution. The primal solution from the primal feasibility phase and dual solution from the dual feasibility phase are then combined (forming a solution of type `POINT_TYPE_FEASIBILITY_POLISHING_SOLUTION`) and checked against the termination criteria.
optional bool use_feasibility_polishing = 30 [default = false];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 12108 of file Solvers.java.
| boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.hasVerbosityLevel | ( | ) |
The verbosity of logging. 0: No informational logging. (Errors are logged.) 1: Summary statistics only. No iteration-level details. 2: A table of iteration-level statistics is logged. (See ToShortString() in primal_dual_hybrid_gradient.cc). 3: A more detailed table of iteration-level statistics is logged. (See ToString() in primal_dual_hybrid_gradient.cc). 4: For iteration-level details, prints the statistics of both the average (prefixed with A) and the current iterate (prefixed with C). Also prints internal algorithmic state and details. Logging at levels 2-4 also includes messages from level 1.
optional int32 verbosity_level = 26 [default = 0];
Implements operations_research.pdlp.Solvers.PrimalDualHybridGradientParamsOrBuilder.
Definition at line 10379 of file Solvers.java.
|
protected |
Definition at line 9458 of file Solvers.java.
| final boolean operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.isInitialized | ( | ) |
Definition at line 9814 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeAdaptiveLinesearchParameters | ( | operations_research.pdlp.Solvers.AdaptiveLinesearchParams | value | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Definition at line 11390 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeFrom | ( | com.google.protobuf.CodedInputStream | input, |
| com.google.protobuf.ExtensionRegistryLite | extensionRegistry ) throws java.io.IOException |
Definition at line 9819 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeFrom | ( | com.google.protobuf.Message | other | ) |
Definition at line 9711 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeFrom | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams | other | ) |
Definition at line 9720 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeMalitskyPockParameters | ( | operations_research.pdlp.Solvers.MalitskyPockParams | value | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Definition at line 11511 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergePresolveOptions | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptions | value | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Definition at line 10951 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeTerminationCriteria | ( | operations_research.pdlp.Solvers.TerminationCriteria | value | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Definition at line 10068 of file Solvers.java.
| final Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.mergeUnknownFields | ( | final com.google.protobuf.UnknownFieldSet | unknownFields | ) |
Definition at line 12235 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setAdaptiveLinesearchParameters | ( | operations_research.pdlp.Solvers.AdaptiveLinesearchParams | value | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Definition at line 11360 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setAdaptiveLinesearchParameters | ( | operations_research.pdlp.Solvers.AdaptiveLinesearchParams.Builder | builderForValue | ) |
optional .operations_research.pdlp.AdaptiveLinesearchParams adaptive_linesearch_parameters = 18;
Definition at line 11376 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setDiagonalQpTrustRegionSolverTolerance | ( | double | value | ) |
The solve tolerance of the experimental trust region solver for diagonal QPs, controlling the accuracy of binary search over a one-dimensional scaling parameter. Smaller values imply smaller relative error of the final solution vector. TODO(user): Find an expression for the final relative error.
optional double diagonal_qp_trust_region_solver_tolerance = 24 [default = 1e-08];
| value | The diagonalQpTrustRegionSolverTolerance to set. |
Definition at line 12047 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setField | ( | com.google.protobuf.Descriptors.FieldDescriptor | field, |
| java.lang.Object | value ) |
Definition at line 9683 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setHandleSomePrimalGradientsOnFiniteBoundsAsResiduals | ( | boolean | value | ) |
See https://developers.google.com/optimization/lp/pdlp_math#treating_some_variable_bounds_as_infinite for a description of this flag.
optional bool handle_some_primal_gradients_on_finite_bounds_as_residuals = 29 [default = true];
| value | The handleSomePrimalGradientsOnFiniteBoundsAsResiduals to set. |
Definition at line 11901 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setInfiniteConstraintBoundThreshold | ( | double | value | ) |
Constraint bounds with absolute value at least this threshold are replaced with infinities. NOTE: This primarily affects the relative convergence criteria. A smaller value makes the relative convergence criteria stronger. It also affects the problem statistics LOG()ed at the start of the run, and the default initial primal weight, since that is based on the norm of the bounds.
optional double infinite_constraint_bound_threshold = 22 [default = inf];
| value | The infiniteConstraintBoundThreshold to set. |
Definition at line 11834 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setInitialPrimalWeight | ( | double | value | ) |
The initial value of the primal weight (i.e., the ratio of primal and dual
step sizes). The primal weight remains fixed throughout the solve if
primal_weight_update_smoothing = 0.0. If unset, the default is the ratio of
the norm of the objective vector to the L2 norm of the combined constraint
bounds vector (as defined above). If this ratio is not finite and positive,
then the default is 1.0 instead. For tuning, try powers of 10, for example,
from 10^{-6} to 10^6.
optional double initial_primal_weight = 8;
| value | The initialPrimalWeight to set. |
Definition at line 10869 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setInitialStepSizeScaling | ( | double | value | ) |
Scaling factor applied to the initial step size (all step sizes if linesearch_rule == CONSTANT_STEP_SIZE_RULE).
optional double initial_step_size_scaling = 25 [default = 1];
| value | The initialStepSizeScaling to set. |
Definition at line 11615 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setL2NormRescaling | ( | boolean | value | ) |
If true, applies L_2 norm rescaling after the Ruiz rescaling. Heuristically this has been found to help convergence.
optional bool l2_norm_rescaling = 10 [default = true];
| value | The l2NormRescaling to set. |
Definition at line 11119 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setLinesearchRule | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.LinesearchRule | value | ) |
Linesearch rule applied at each major iteration.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.LinesearchRule linesearch_rule = 12 [default = ADAPTIVE_LINESEARCH_RULE];
| value | The linesearchRule to set. |
Definition at line 11312 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setLInfRuizIterations | ( | int | value | ) |
Number of L_infinity Ruiz rescaling iterations to apply to the constraint matrix. Zero disables this rescaling pass. Recommended values to try when tuning are 0, 5, and 10.
optional int32 l_inf_ruiz_iterations = 9 [default = 5];
| value | The lInfRuizIterations to set. |
Definition at line 11058 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setLogIntervalSeconds | ( | double | value | ) |
Time between iteration-level statistics logging (if `verbosity_level > 1`). Since iteration-level statistics are only generated when performing termination checks, logs will be generated from next termination check after `log_interval_seconds` have elapsed. Should be >= 0.0. 0.0 (the default) means log statistics at every termination check.
optional double log_interval_seconds = 31 [default = 0];
| value | The logIntervalSeconds to set. |
Definition at line 10501 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setMajorIterationFrequency | ( | int | value | ) |
The frequency at which extra work is performed to make major algorithmic decisions, e.g., performing restarts and updating the primal weight. Major iterations also trigger a termination check. For best performance using the NO_RESTARTS or EVERY_MAJOR_ITERATION rule, one should perform a log-scale grid search over this parameter, for example, over powers of two. ADAPTIVE_HEURISTIC is mostly insensitive to this value.
optional int32 major_iteration_frequency = 4 [default = 64];
| value | The majorIterationFrequency to set. |
Definition at line 10576 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setMalitskyPockParameters | ( | operations_research.pdlp.Solvers.MalitskyPockParams | value | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Definition at line 11481 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setMalitskyPockParameters | ( | operations_research.pdlp.Solvers.MalitskyPockParams.Builder | builderForValue | ) |
optional .operations_research.pdlp.MalitskyPockParams malitsky_pock_parameters = 19;
Definition at line 11497 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setNecessaryReductionForRestart | ( | double | value | ) |
For ADAPTIVE_HEURISTIC only: A relative reduction in the potential function by this amount triggers a restart if, additionally, the quality of the iterates appears to be getting worse. The value must be in the interval [sufficient_reduction_for_restart, 1). Smaller values make restarts less frequent, and larger values make them more frequent.
optional double necessary_reduction_for_restart = 17 [default = 0.9];
| value | The necessaryReductionForRestart to set. |
Definition at line 11252 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setNumShards | ( | int | value | ) |
For more efficient parallel computation, the matrices and vectors are divided (virtually) into num_shards shards. Results are computed independently for each shard and then combined. As a consequence, the order of computation, and hence floating point roundoff, depends on the number of shards so reproducible results require using the same value for num_shards. However, for efficiency num_shards should a be at least num_threads, and preferably at least 4*num_threads to allow better load balancing. If num_shards is positive, the computation will use that many shards. Otherwise a default that depends on num_threads will be used.
optional int32 num_shards = 27 [default = 0];
| value | The numShards to set. |
Definition at line 10261 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setNumThreads | ( | int | value | ) |
The number of threads to use. Must be positive. Try various values of num_threads, up to the number of physical cores. Performance may not be monotonically increasing with the number of threads because of memory bandwidth limitations.
optional int32 num_threads = 2 [default = 1];
| value | The numThreads to set. |
Definition at line 10178 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setPresolveOptions | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptions | value | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Definition at line 10921 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setPresolveOptions | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.PresolveOptions.Builder | builderForValue | ) |
optional .operations_research.pdlp.PrimalDualHybridGradientParams.PresolveOptions presolve_options = 16;
Definition at line 10937 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setPrimalWeightUpdateSmoothing | ( | double | value | ) |
This parameter controls exponential smoothing of log(primal_weight) when a primal weight update occurs (i.e., when the ratio of primal and dual step sizes is adjusted). At 0.0, the primal weight will be frozen at its initial value and there will be no dynamic updates in the algorithm. At 1.0, there is no smoothing in the updates. The default of 0.5 generally performs well, but has been observed on occasion to trigger unstable swings in the primal weight. We recommend also trying 0.0 (disabling primal weight updates), in which case you must also tune initial_primal_weight.
optional double primal_weight_update_smoothing = 7 [default = 0.5];
| value | The primalWeightUpdateSmoothing to set. |
Definition at line 10788 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setRandomProjectionSeeds | ( | int | index, |
| int | value ) |
Seeds for generating (pseudo-)random projections of iterates during termination checks. For each seed, the projection of the primal and dual solutions onto random planes in primal and dual space will be computed and added the IterationStats if record_iteration_stats is true. The random planes generated will be determined by the seeds, the primal and dual dimensions, and num_threads.
repeated int32 random_projection_seeds = 28 [packed = true];
| index | The index to set the value at. |
| value | The randomProjectionSeeds to set. |
Definition at line 11711 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setRecordIterationStats | ( | boolean | value | ) |
If true, the iteration_stats field of the SolveLog output will be populated at every iteration. Note that we only compute solution statistics at termination checks. Setting this parameter to true may substantially increase the size of the output.
optional bool record_iteration_stats = 3;
| value | The recordIterationStats to set. |
Definition at line 10334 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setRepeatedField | ( | com.google.protobuf.Descriptors.FieldDescriptor | field, |
| int | index, | ||
| java.lang.Object | value ) |
Definition at line 9699 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setRestartStrategy | ( | operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.RestartStrategy | value | ) |
NO_RESTARTS and EVERY_MAJOR_ITERATION occasionally outperform the default. If using a strategy other than ADAPTIVE_HEURISTIC, you must also tune major_iteration_frequency.
optional .operations_research.pdlp.PrimalDualHybridGradientParams.RestartStrategy restart_strategy = 6 [default = ADAPTIVE_HEURISTIC];
| value | The restartStrategy to set. |
Definition at line 10707 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setSufficientReductionForRestart | ( | double | value | ) |
For ADAPTIVE_HEURISTIC and ADAPTIVE_DISTANCE_BASED only: A relative reduction in the potential function by this amount always triggers a restart. Must be between 0.0 and 1.0.
optional double sufficient_reduction_for_restart = 11 [default = 0.1];
| value | The sufficientReductionForRestart to set. |
Definition at line 11182 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setTerminationCheckFrequency | ( | int | value | ) |
The frequency (based on a counter reset every major iteration) to check for termination (involves extra work) and log iteration stats. Termination checks do not affect algorithmic progress unless termination is triggered.
optional int32 termination_check_frequency = 5 [default = 64];
| value | The terminationCheckFrequency to set. |
Definition at line 10643 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setTerminationCriteria | ( | operations_research.pdlp.Solvers.TerminationCriteria | value | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Definition at line 10038 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setTerminationCriteria | ( | operations_research.pdlp.Solvers.TerminationCriteria.Builder | builderForValue | ) |
optional .operations_research.pdlp.TerminationCriteria termination_criteria = 1;
Definition at line 10054 of file Solvers.java.
| final Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setUnknownFields | ( | final com.google.protobuf.UnknownFieldSet | unknownFields | ) |
Definition at line 12229 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setUseDiagonalQpTrustRegionSolver | ( | boolean | value | ) |
When solving QPs with diagonal objective matrices, this option can be turned on to enable an experimental solver that avoids linearization of the quadratic term. The `diagonal_qp_solver_accuracy` parameter controls the solve accuracy. TODO(user): Turn this option on by default for quadratic programs after numerical evaluation.
optional bool use_diagonal_qp_trust_region_solver = 23 [default = false];
| value | The useDiagonalQpTrustRegionSolver to set. |
Definition at line 11974 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setUseFeasibilityPolishing | ( | boolean | value | ) |
If true, periodically runs feasibility polishing, which attempts to move from latest average iterate to one that is closer to feasibility (i.e., has smaller primal and dual residuals) while probably increasing the objective gap. This is useful primarily when the feasibility tolerances are fairly tight and the objective gap tolerance is somewhat looser. Note that this does not change the termination criteria, but rather can help achieve the termination criteria more quickly when the objective gap is not as important as feasibility. `use_feasibility_polishing` cannot be used with glop presolve, and requires `handle_some_primal_gradients_on_finite_bounds_as_residuals == false`. `use_feasibility_polishing` can only be used with linear programs. Feasibility polishing runs two separate phases, primal feasibility and dual feasibility. The primal feasibility phase runs PDHG on the primal feasibility problem (obtained by changing the objective vector to all zeros), using the average primal iterate and zero dual (which is optimal for the primal feasibility problem) as the initial solution. The dual feasibility phase runs PDHG on the dual feasibility problem (obtained by changing all finite variable and constraint bounds to zero), using the average dual iterate and zero primal (which is optimal for the dual feasibility problem) as the initial solution. The primal solution from the primal feasibility phase and dual solution from the dual feasibility phase are then combined (forming a solution of type `POINT_TYPE_FEASIBILITY_POLISHING_SOLUTION`) and checked against the termination criteria.
optional bool use_feasibility_polishing = 30 [default = false];
| value | The useFeasibilityPolishing to set. |
Definition at line 12182 of file Solvers.java.
| Builder operations_research.pdlp.Solvers.PrimalDualHybridGradientParams.Builder.setVerbosityLevel | ( | int | value | ) |
The verbosity of logging. 0: No informational logging. (Errors are logged.) 1: Summary statistics only. No iteration-level details. 2: A table of iteration-level statistics is logged. (See ToShortString() in primal_dual_hybrid_gradient.cc). 3: A more detailed table of iteration-level statistics is logged. (See ToString() in primal_dual_hybrid_gradient.cc). 4: For iteration-level details, prints the statistics of both the average (prefixed with A) and the current iterate (prefixed with C). Also prints internal algorithmic state and details. Logging at levels 2-4 also includes messages from level 1.
optional int32 verbosity_level = 26 [default = 0];
| value | The verbosityLevel to set. |
Definition at line 10423 of file Solvers.java.
1.10.0