Class TrainedDagSimulatorGNM2.Params
java.lang.Object
edu.cmu.tetrad.sem.TrainedDagSimulatorGNM2.Params
- Enclosing class:
TrainedDagSimulatorGNM2
Encapsulates configuration parameters for the TrainedDagSimulatorGNM class.
This class defines a set of tunable options that control the behavior of the simulator,
such as training parameters, random seed initialization, and simulation strategies.
These parameters can be customized to achieve different trade-offs in simulation fidelity,
efficiency, or interpretability.
-
Field Summary
FieldsModifier and TypeFieldDescriptionintBatch size for the simulator.booleanWhether to preserve marginal distributions at root nodes during simulation.intNumber of training epochs for the simulator.intConfiguration parameters for the TrainedDagSimulatorGNM class.doubleL2 regularization parameter for the simulator.doubleInterpolation factor for unconditional vs structural mechanism during simulation.doubleLearning rate for the simulator.intMaximum number of discrete levels to consider for simulation.intMaximum number of residual bootstrap strata to consider.longRandom seed for the simulator.booleanWhether to stratify residual bootstrap by discrete-parent signature.doubleA threshold value used to issue a warning for the "z" score in statistical or analytical calculations.doubleA threshold value used to issue a warning for the "z" score in statistical or analytical calculations. -
Constructor Summary
Constructors -
Method Summary
-
Field Details
-
epochs
public int epochsNumber of training epochs for the simulator. This parameter controls the number of iterations over the entire dataset during training. A higher number of epochs can lead to more refined models but may also increase training time. -
lr
public double lrLearning rate for the simulator. This parameter controls the step size at each iteration during training. A smaller learning rate can lead to more stable convergence but may require more epochs. -
l2
public double l2L2 regularization parameter for the simulator. This parameter controls the amount of L2 regularization applied during training. A higher value can lead to more regularization and potentially prevent overfitting. -
batchSize
public int batchSizeBatch size for the simulator. This parameter controls the number of samples processed in each training iteration. A larger batch size can lead to faster training but may require more memory. -
seed
public long seedRandom seed for the simulator. This parameter initializes the random number generator used for various operations, ensuring reproducibility of simulation results when the same seed is used. -
maxDiscreteLevels
public int maxDiscreteLevelsMaximum number of discrete levels to consider for simulation. This parameter controls the maximum number of discrete levels allowed for variables in the simulation. Variables with more levels than this threshold will be treated as continuous. -
bootstrapRoots
public boolean bootstrapRootsWhether to preserve marginal distributions at root nodes during simulation. This parameter controls whether the simulator should maintain the marginal distributions of root nodes (nodes with no parents) during the simulation process. When set to true, the simulator will attempt to preserve these distributions, which can be useful for certain types of simulations or analyses. -
lambdaParents
public double lambdaParentsInterpolation factor for unconditional vs structural mechanism during simulation. This parameter controls the balance between using the structural mechanism (based on parent variables) and the unconditional mechanism (ignoring parent variables) when simulating the effect of a variable. A value of 1.0 means fully structural, while 0.0 means fully unconditional. -
zWarn1
public double zWarn1A threshold value used to issue a warning for the "z" score in statistical or analytical calculations. This variable may represent a critical boundary, signaling the occurrence of a significant event or aiding in anomaly detection. The default value is set to 4.0. -
zWarn2
public double zWarn2A threshold value used to issue a warning for the "z" score in statistical or analytical calculations. This variable may represent a critical boundary, signaling the occurrence of a significant event or aiding in anomaly detection. The default value is set to 6.0. -
stratifyResidualsByDiscreteParents
public boolean stratifyResidualsByDiscreteParentsWhether to stratify residual bootstrap by discrete-parent signature. This parameter controls whether the simulator should stratify residual bootstrap based on the discrete-parent signature of variables. Stratification can help improve the accuracy of simulations in mixed variable settings. -
maxResidualStrata
public int maxResidualStrataMaximum number of residual bootstrap strata to consider. This parameter controls the maximum number of strata maps allowed for residual bootstrap in large categorical parent sets. It helps prevent excessive memory usage and computational complexity.
-
Constructor Details
-
Params
public Params()Default constructor for the Params class. Initializes an instance of Params with default values for all fields.
-