Parameters used in the LP process. More...

#include <BCP_lp_param.hpp>

Inheritance diagram for BCP_lp_par:

Public Types
enum	chr_params { BranchOnCuts, CompareNewCutsToOldOnes, CompareNewVarsToOldOnes, DoReducedCostFixingAtZero, DoReducedCostFixingAtAnything, MessagePassingIsSerial, ReportWhenDefaultIsExecuted, SendFathomedNodeDesc, NoCompressionAtFathom, UseExplicitStorage, Lp_DumpNodeDescCuts, Lp_DumpNodeDescVars, LpVerb_First, LpVerb_AddedCutCount, LpVerb_AddedVarCount, LpVerb_ChildrenInfo, LpVerb_ColumnGenerationInfo, LpVerb_CutsToCutPoolCount, LpVerb_VarsToVarPoolCount, LpVerb_FathomInfo, LpVerb_IterationCount, LpVerb_RelaxedSolution, LpVerb_FinalRelaxedSolution, LpVerb_LpMatrixSize, LpVerb_LpSolutionValue, LpVerb_MatrixCompression, LpVerb_NodeTime, LpVerb_PresolvePositions, LpVerb_PresolveResult, LpVerb_ProcessedNodeIndex, LpVerb_ReportCutGenTimeout, LpVerb_ReportVarGenTimeout, LpVerb_ReportLocalCutPoolSize, LpVerb_ReportLocalVarPoolSize, LpVerb_RepricingResult, LpVerb_VarTightening, LpVerb_RowEffectivenessCount, LpVerb_StrongBranchPositions, LpVerb_StrongBranchResult, LpVerb_GeneratedCutCount, LpVerb_GeneratedVarCount, LpVerb_Last, end_of_chr_params }
	Character parameters. More...

enum	int_params { NiceLevel, ScaleMatrix, SlackCutDiscardingStrategy, CutEffectiveCountBeforePool, CutPoolCheckFrequency, VarPoolCheckFrequency, IneffectiveConstraints, IneffectiveBeforeDelete, MaxNonDualFeasToAdd_Min, MaxNonDualFeasToAdd_Max, CutViolationNorm, MaxCutsAddedPerIteration, MaxVarsAddedPerIteration, MaxLeftoverCutNum, DeletedColToCompress_Min, DeletedRowToCompress_Min, MaxPresolveIter, StrongBranchNum, StrongBranch_CloseToHalfNum, BranchingObjectComparison, ChildPreference, FeasibilityTest, WarmstartInfo, InfoForCG, InfoForVG, end_of_int_params }
	Integer parameters. More...

enum	dbl_params { Granularity, DeletedColToCompress_Frac, DeletedRowToCompress_Frac, MaxNonDualFeasToAdd_Frac, MaxLeftoverCutFrac, IntegerTolerance, FirstLP_FirstCutTimeout, LaterLP_FirstCutTimeout, FirstLP_AllCutsTimeout, LaterLP_AllCutsTimeout, FirstLP_FirstVarTimeout, LaterLP_FirstVarTimeout, FirstLP_AllVarsTimeout, LaterLP_AllVarsTimeout, MaxRunTime, end_of_dbl_params }
	Double parameters. More...

enum	str_params { LogFileName, end_of_str_params }
	String parameters. More...

enum	str_array_params { str_array_dummy, end_of_str_array_params }
	There are no string array parameters. More...

Verbosity Parameters.
These are all true/false parameters; if the parameter value is set then the corresponding information is printed.

Detailed Description

Parameters used in the LP process.

These parameters can be set in the original parameter file by including the following line:
BCP_{parameter name} {parameter value}.

Definition at line 14 of file BCP_lp_param.hpp.

Member Enumeration Documentation

enum BCP_lp_par::chr_params

Character parameters.

All of these variables are used as booleans (true = 1, false = 0).

Enumerator
BranchOnCuts	If true, BCP supports branching on cuts by providing potential branching candidates for the user. These are cuts that were added to the formulation at some point but became slack later in subsequent LP relaxations. Values: true (1), false (0). Default: 0.
CompareNewCutsToOldOnes	If true then the LP process will check each newly received cut whether it already exists in the local cut pool or not. Values: true (1), false (0). Default: 1.
CompareNewVarsToOldOnes	If true then the LP process will check each newly arrived variable whether it already exists in the local variable pool or not. Values: true (1), false (0). Default: 1.
DoReducedCostFixingAtZero	If true the BCP will attempt to do reduced cost fixing only for variables currently at zero. Values: true (1), false (0). Default: 1.
DoReducedCostFixingAtAnything	If true the BCP will attempt to do reduced cost fixing for any variable, no matter what is their current value. Values: true (1), false (0). Default: 1.
MessagePassingIsSerial	Indicates whether message passing is serial (all processes are on the same processor) or not. Values: true (1), false (0). Default: 1.
ReportWhenDefaultIsExecuted	Print out a message when the default version of an overridable method is executed. Default: 1.
SendFathomedNodeDesc	Whether to send back the description of fathomed search tree nodes to the Tree Manager. (Might be needed to write proof of optimality.) Values: true (1), false (0). Default: 1.
NoCompressionAtFathom	Whether we should refrain from compressing the problem description right before a fathomed node's description is sent back to the tree manager. Values: true (1), false (0). Default: 0.
UseExplicitStorage
Lp_DumpNodeDescCuts
Lp_DumpNodeDescVars
LpVerb_First	Just a marker for the first LpVerb.
LpVerb_AddedCutCount	Print the number of cuts added from the local cut pool in the current iteration. (BCP_lp_main_loop)
LpVerb_AddedVarCount	Print the number of variables added from the local variable pool in the curent iteration. (BCP_lp_main_loop)
LpVerb_ChildrenInfo	After a branching object is selected print what happens to the presolved children (e.g., fathomed). (BCP_print_brobj_stat)
LpVerb_ColumnGenerationInfo	Print the number of variables generated before resolving the Lp ir fathoming a node. (BCP_lp_fathom)
LpVerb_CutsToCutPoolCount	Print the number of cuts sent from the LP to the cut pool. (BCP_lp_send_cuts_to_cp)
LpVerb_VarsToVarPoolCount	** This parameter is not used anywhere!!! **
LpVerb_FathomInfo	Print information related to fathoming. (BCP_lp_main_loop, BCP_lp_perform_fathom, BCP_lp_branch) (BCP_lp_fathom)
LpVerb_IterationCount	Print the "Starting iteration x" line. (BCP_lp_main_loop)
LpVerb_RelaxedSolution	Turn on the user hook "display_lp_solution". (BCP_lp_main_loop)
LpVerb_FinalRelaxedSolution	Turn on the user hook "display_lp_solution" for the last LP relaxation solved at a search tree node. (BCP_lp_main_loop)
LpVerb_LpMatrixSize	** This parameter is not used anywhere!!! **
LpVerb_LpSolutionValue	Print the size of the problem matrix and the LP solution value after resolving the LP. (BCP_lp_main_loop)
LpVerb_MatrixCompression	Print the number of columns and rows that were deleted during matrix compression. (BCP_lp_delete_cols_and_rows)
LpVerb_NodeTime	For each tree node print out how much time was spent on it.
LpVerb_PresolvePositions	Print detailed information about all the branching candidates during strong branching. LpVerb_PresolveResult must be set for this parameter to have an effect. (BCP_lp_perform_strong_branching)
LpVerb_PresolveResult	Print information on the presolved branching candidates during strong branching. (BCP_lp_perform_strong_branching)
LpVerb_ProcessedNodeIndex	Print the "Processing NODE x on LEVEL y" line. (BCP_lp-main_loop)
LpVerb_ReportCutGenTimeout	Print information if receiving cuts is timed out. (BCP_lp_generate_cuts)
LpVerb_ReportVarGenTimeout	Print information if receiving variables is timed out. (BCP_lp_generate_vars)
LpVerb_ReportLocalCutPoolSize	Print the current number of cuts in the cut pool. This number is printed several times: before and after generating columns at the current iteration, after removing non-essential cuts, etc. (BCP_lp_generate_cuts)
LpVerb_ReportLocalVarPoolSize	Similar as above for variables. (BCP_lp_generate_vars)
LpVerb_RepricingResult	** This parameter is not used anywhere!!! **
LpVerb_VarTightening	Print the number of variables whose bounds have been changed by reduced cost fixing or logical fixing. (BCP_lp_fix_vars)
LpVerb_RowEffectivenessCount	Print the number of ineffective rows in the current problem. The definition of what rows are considered ineffective is determined by the paramter IneffectiveConstraints. (BCP_lp_adjust_row_effectiveness)
LpVerb_StrongBranchPositions	Print detailed information on the branching candidate selected by strong branching. LpVerb_StrongBranchResult must be set fo this parameter to have an effect. (BCP_print_brobj_stat)
LpVerb_StrongBranchResult	Print information on the branching candidate selected by strong branching. (BCP_print_brobj_stat)
LpVerb_GeneratedCutCount	Print the number of cuts generated during this iteration (since the LP was resolved last time). (BCP_lp_main_loop)
LpVerb_GeneratedVarCount	Print the number of variables generated during this iteration. (BCP_lp_main_loop)
LpVerb_Last	Just a marker for the last LpVerb.
end_of_chr_params

Definition at line 17 of file BCP_lp_param.hpp.

enum BCP_lp_par::int_params

Integer parameters.

Enumerator
NiceLevel	What should be the "niceness" of the LP process. In a *nix environment the LP process will be reniced to this level. Values: whatever the operating system accepts. Default: 0.
ScaleMatrix	Indicates how matrix scaling should be performed. This parameter is directly passed to the LP solver's `load_lp` member function. Values: whatever is valid for the LP solver that the user has passed on to BCP. Default: 0.
SlackCutDiscardingStrategy	The slack cut discarding strategy used in the default version of the function `purge_slack_pool()`. Values: BCP_slack_cut_discarding. Default: `BCP_DiscardSlackCutsAtNewIteration`
CutEffectiveCountBeforePool	A cut has to remain effective through this many iterations in the LP before it is sent to the Cut Pool process. The default 1000 effectively says that only those cuts are sent to the pool which are effective at branching. Values: any positive number. Default: 1000.
CutPoolCheckFrequency	The Cut Pool is queried for violated valid inequalities after the first LP relaxation is solved and then after every this many iterations. Values: any positive number. Default: 10.
VarPoolCheckFrequency	The Variable Pool is queried for columns that improve the formulation after the first LP realxation is solved, and then after every this many iterations. Values: any positive number. Default: 10.
IneffectiveConstraints	Indicates which constraints should be considered ineffective. Values: BCP_IneffectiveConstraints. Default: `BCP_IneffConstr_ZeroDualValue`.
IneffectiveBeforeDelete	How many times in a row a constraint must be found ineffective before it is marked for deletion. Values: any positive number. Default: 1.
MaxNonDualFeasToAdd_Min	The number of non dual-feasible colums that can be added at a time to the formulation is a certain fraction (`MaxNonDualFeasToAdd_Frac`) of the number of columns in the current formulation. However, if the computed number is outside of the range `[MaxNonDualFeasToAdd_Min, MaxNonDualFeasToAdd_Max]` then it will be set to the appropriate endpoint of the range. Values: . Default: 5.
MaxNonDualFeasToAdd_Max	See the description of the previous parameter. Values: . Default: 200.
CutViolationNorm	How cut violation should be computed. Values: BCP_CutViolationNorm. Default: `BCP_CutViolationNorm_Plain` (violation in the usual sense). (The other option is `BCP_CutViolationNorm_Distance`: the distance of the fractional point from the cut.
MaxCutsAddedPerIteration	The maximum number of violated valid inequalities that can be added per iteration. Values: . Default: 100,000.
MaxVarsAddedPerIteration	The maximum number of variables that can be added per iteration. Values: . Default: 100,000.
MaxLeftoverCutNum	The maximum number of violated but not added cuts to be kept from one iteration to the next. Also see the MaxLeftoverCutFrac parameter.
DeletedColToCompress_Min	The number of columns that must be marked for deletion before matrix compression can occur. Matrix compressions also subject to a minimum number of marked columns as a fraction of the current number of columns (see `DeletedColToCompress_Frac`). Values: positive integer. Default: 10.
DeletedRowToCompress_Min	The number of rows that must be marked for deletion before matrix compression can occur. Matrix compressionis also subject to a minimum number of marked columns as a fraction of the current number of columns (see `DeletedRowToCompress_Frac`). Values: positive integer. Default: 10.
MaxPresolveIter	Upper limit on the number of iterations performed in each of the children of the search tree node when presolving branching candidates. This parameter is passed onto the LP solver. If the parameter is set to -1 then the branching candidates are not presolved and the first branching candidate is chosen (if there is any). Note that for different LP solvers (e.g., simplex based algorithm or the Volume Algorithm) the meaning of an iteration is different, thus this parameter will be set differently. Values: . Default: 100,000.
StrongBranchNum	Specifies how many branching variables with values close to half between two integers should be chosen by the built-in branching variable selection routine `select_branching_candidates()` (if this routine is used at all). Values: . Default: 3.
StrongBranch_CloseToHalfNum	<bf>Deprecated parameter. Use StrongBranchNum</bf>
BranchingObjectComparison	Specifies the rule used for built-in branching object comparison (if the buit-in routine is used at all). Values: BCP_branching_object_comparison. Default: `BCP_HighestLowObjval`.
ChildPreference	Specifies the rule used for selecting one of the children of the search tree node for diving. Values: BCP_child_preference. Default: `BCP_PreferChild_LowBound`.
FeasibilityTest	Specifies which built-in MIP feasibility testing routine should be invoked (if a buit-in routine is used at all). Values: BCP_feasibility_test. Default: `BCP_FullTest_Feasible`
WarmstartInfo	Specifies how warmstart information should be stored in the TM. Possible valueas: BCP_WarmstartNone, BCP_WarmstartRoot, BCP_WarmstartParent. The first is obvious, the second means the root node's warmstart info should be sent out along with each node, the last means the parent's warmstart info should be sent out. Default: BCP_WarmstartParent
InfoForCG	Indicates what part of the primal solution is sent to the Cut Generator process if the BCP_lp_user::pack_primal_solution() method is not overridden. Values: See BCP_primal_solution_description. Default: `BCP_PrimalSolution_Nonzeros`.
InfoForVG	Indicates what part of the dual solution is sent to the Variable Generator process if the BCP_lp_user::pack_dual_solution() method is not overridden. Values: See BCP_dual_solution_description. Default: `BCP_DualSolution_Full`.
end_of_int_params

Definition at line 163 of file BCP_lp_param.hpp.

enum BCP_lp_par::dbl_params

Double parameters.

Enumerator
Granularity	The minimum difference between the objective value of any two feasible solution (with different objective values). Values: . Default: .
DeletedColToCompress_Frac	The fraction of columns that must be marked for deletion before matrix compression can occur. Matrix compression is also subject to a minimum number of marked columns (see `DeletedColToCompress_Min`). Values: . Default: `1e-8`.
DeletedRowToCompress_Frac	The fraction of rows that must be marked for deletion before matrix compression can occur. Matrix compression is also subject to a minimum number of marked rows (see `DeletedRowToCompress_Min`). Values: . Default: `1e-8`.
MaxNonDualFeasToAdd_Frac	The number of non dual-feasible colums that can be added to the formulation at a time cannot exceed this fraction of the the number of columns in the curent formulation. However, this limit can be overruled by the hard bounds `MaxNonDualFeasToAdd_Min` and `MaxNonDualFeasToAdd_Max` (see above). Default: 0.05.
MaxLeftoverCutFrac	The maximum fraction of the violated but not added cuts to be kept from one iteration to the next. Also see the MaxLeftoverCutNum parameter.
IntegerTolerance	Values not further from an integer value than the value of this parameter are considered to be integer. Values: . Default: .
FirstLP_FirstCutTimeout	This and the following three parameters control how long the LP process waits for generated cuts. The parameters specify waiting times (in seconds) separately for the first and later LP relaxations at a search tree node, and also the time to receive the first cut vs all the cuts. Note that it might make sense to set `AllCutsTimeout` for a shorter time than the first cut time out: "We are willing to wait 5 seconds to receive a cut, but if we do receive a cut the total time we wait is only 2 seconds." This parameter specifies the time to wait for the first generated cut at the first LP relaxation at a search tree node. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
LaterLP_FirstCutTimeout	This parameter specifies the time to wait for the first generated cut at iterations that are not the first at a search tree node. See the remarks at `FirstLP_FirstCutTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
FirstLP_AllCutsTimeout	This parameter specifies the time to wait for cuts at the first LP relaxation at a search tree node. See the remarks at `FirstLP_FirstCutTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
LaterLP_AllCutsTimeout	This parameter specifies the time to wait for cuts at iterations that are not the first at a search tree node. See the remarks at `FirstLP_FirstCutTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
FirstLP_FirstVarTimeout	This and the following three parameters control how long the LP process waits for generated variables. These parameters are analogous to the ones that control the waiting time for generated cuts. See the remarks at `FirstLP_FirstCutTimeout` as well. This parameter specifies the time to wait for the first generated variable at the first LP relaxation at a search tree node. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
LaterLP_FirstVarTimeout	This parameter specifies the time to wait for the first generated variable at iterations that are not the first at a search tree node. See the remarks at `FirstLP_FirstVarTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
FirstLP_AllVarsTimeout	This parameter specifies the time to wait for variables at the first LP relaxation at a search tree node. See the remarks at `FirstLP_FirstVarTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
LaterLP_AllVarsTimeout	This parameter specifies the time to wait for variables at iterations that are not the first at a search tree node. See the remarks at `FirstLP_FirstVarTimeout` as well. Values: Any non-negative number represents a waiting time in seconds. Negative numbers indicate no time limit on waiting. Default: -1.
MaxRunTime	Maximum allowed running time.
end_of_dbl_params

Definition at line 313 of file BCP_lp_param.hpp.

enum BCP_lp_par::str_params

String parameters.

Enumerator
LogFileName	The filename all the output should go to.
end_of_str_params

Definition at line 424 of file BCP_lp_param.hpp.

enum BCP_lp_par::str_array_params

There are no string array parameters.

Enumerator
str_array_dummy
end_of_str_array_params

Definition at line 432 of file BCP_lp_param.hpp.

The documentation for this struct was generated from the following file:

/tmp/Bcp-1.4.4/Bcp/src/include/BCP_lp_param.hpp

Public Types

Detailed Description

Member Enumeration Documentation