Class NodeCVSummary

java.lang.Object

edu.cmu.tetrad.sem.NodeCVSummary

All Implemented Interfaces:

TetradSerializable, Serializable

public final class NodeCVSummary
extends Object
implements TetradSerializable

Immutable per-node summary produced by k-fold cross-validation in NNEstimator#crossValidate(int).

For continuous nodes the primary OOS metric is oosMse and the derived oosR2 (improvement over predicting the marginal mean). For discrete nodes the primary metric is oosXent (average cross-entropy on held-out rows).

Root nodes (no parents) have no conditional prediction and carry Double.NaN for all prediction metrics.

See Also:

• Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
final double	baselineMse	Baseline MSE: variance of the child variable estimated from the full dataset.
final double	baselineXent	Baseline cross-entropy: entropy of the empirical marginal distribution of the child variable.
final boolean	discreteChild	true if the variable is discrete.
final String	node	Variable name.
final int	numFolds	Number of folds used.
final double	oosMse	Mean OOS MSE across folds: average of (observed - predicted)² on held-out rows.
final double	oosR2	OOS R²: 1 - oosMse / baselineMse.
final double	oosXent	Mean OOS cross-entropy across folds on held-out rows.
final List<String>	parents	Parent variable names (empty for root nodes).

Method Summary

Modifier and Type	Method	Description
String	toString()

Methods inherited from class Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Details

node

public final String node

Variable name.

discreteChild

public final boolean discreteChild

true if the variable is discrete.

parents

public final List<String> parents

Parent variable names (empty for root nodes).

numFolds

public final int numFolds

Number of folds used.

oosMse

public final double oosMse

Mean OOS MSE across folds: average of (observed - predicted)² on held-out rows. Uses zero-noise prediction (conditional mean estimate). NaN for discrete or root nodes.

baselineMse

public final double baselineMse

Baseline MSE: variance of the child variable estimated from the full dataset. This is the MSE you would get by always predicting the marginal mean.

oosR2

public final double oosR2

OOS R²: 1 - oosMse / baselineMse. Positive means the NN beats the mean baseline out of sample. NaN if either component is NaN.

oosXent

public final double oosXent

Mean OOS cross-entropy across folds on held-out rows. NaN for continuous or root nodes.

baselineXent

public final double baselineXent

Baseline cross-entropy: entropy of the empirical marginal distribution of the child variable. This is the cross-entropy you would get by always predicting the marginal class probabilities.

Method Details

toString

public String toString()

Overrides:

toString in class Object