Package edu.cmu.tetrad.search

Class TwoStep

java.lang.Object
edu.cmu.tetrad.search.TwoStep
@Deprecated public final class TwoStep extends Object
Deprecated.
Two-Step algorithm (linear, non-Gaussian disturbances; cycles allowed). Orientation: B[i,j] != 0 means edge j -> i.

Step 1: Sparse mask via (Adaptive) Lasso on Xi ~ X\i (Gram form over covariance). Penalty uses paper-style scaling when normalizeLossByN=true: lambda * (ln N)/2. Step 2: ICA (FastICA) to get A, align (Hungarian) so diag(A)=1, B0 = I - A^{-1}. Project to mask, refine with constrained OLS (covariance), pairwise BIC cleanup, then strict two-cycle breaker (optional).

  • Nested Class Summary

    Nested Classes
    Modifier and Type
    Class
    Description
    static final class 
    TwoStep.Result
    Deprecated.
    The Result class encapsulates the output of the TwoStep algorithm's search process.

  • Constructor Summary

    Constructors
    Constructor
    Description
    TwoStep()
    Deprecated.
    Default constructor for the TwoStep class.

  • Method Summary

    Modifier and Type
    Method
    Description
    org.ejml.simple.SimpleMatrix
    getLastA()
    Deprecated.
    Retrieves the last calculated matrix A.
    static org.ejml.simple.SimpleMatrix
    maskFromUndirected(Graph skeleton, List<Node> vars)
    Deprecated.
    Constructs a mask matrix from an undirected graph, indicating the adjacency relationships between a given list of nodes.
    TwoStep.Result
    search(DataSet data)
    Deprecated.
    Performs a search operation over the given dataset to analyze dependencies, refine coefficients, and build a directed graph representation.
    void
    setAdaptiveGamma(double gamma)
    Deprecated.
    Sets the value of the adaptive gamma parameter used in the two-step regression process.
    void
    setBreakTwoCyclesEnabled(boolean enabled)
    Deprecated.
    Enables or disables the breaking of two-cycle constraints in the processing.
    void
    setCoefThreshold(double t)
    Deprecated.
    Sets the coefficient threshold value for the regression process.
    void
    setCondWarnThreshold(double thr)
    Deprecated.
    Sets the condition warning threshold value used in the processing.
    void
    setExternalMask(org.ejml.simple.SimpleMatrix mask)
    Deprecated.
    Sets an external mask for use in the current object.
    void
    setFastIcaAlgorithm(int alg)
    Deprecated.
    Sets the FastICA algorithm type to be used in the Independent Component Analysis (ICA) procedure.
    void
    setFastIcaAlpha(double a)
    Deprecated.
    Sets the alpha parameter for the FastICA (Independent Component Analysis) process.
    void
    setFastIcaFunction(int f)
    Deprecated.
    Sets the FastICA function type to be used within the Independent Component Analysis (ICA) process.
    void
    setFastIcaRowNorm(boolean flag)
    Deprecated.
    Sets whether row normalization is applied in the FastICA (Independent Component Analysis) process.
    void
    setIcaMaxIter(int n)
    Deprecated.
    Sets the maximum number of iterations for the ICA (Independent Component Analysis) process.
    void
    setIcaTol(double t)
    Deprecated.
    Sets the tolerance value for the ICA (Independent Component Analysis) process.
    void
    setLambda(double lambda)
    Deprecated.
    Sets the value of the regularization parameter lambda.
    void
    setMaskThreshold(double t)
    Deprecated.
    Sets the mask threshold value used in the two-step regression process.
    void
    setNormalizeLossByN(boolean flag)
    Deprecated.
    If true, Step-1 uses lambdaEff = lambda * (ln N)/2 (paper-style).
    void
    setRandomSeed(long seed)
    Deprecated.
    Sets the random seed for reproducibility in the algorithm's processes.
    void
    setTwoCycleMinAbs(double minAbs)
    Deprecated.
    Sets the minimum absolute value threshold for breaking two-cycle constraints in the processing.
    void
    setUseAdaptiveLasso(boolean flag)
    Deprecated.
    Sets whether the adaptive lasso method should be used in the two-step regression process.
    void
    setVerbose(boolean v)
    Deprecated.
    Sets whether verbose output is enabled for the process.

    Methods inherited from class java.lang.Object

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

  • Constructor Details

    • TwoStep

      public TwoStep()
      Deprecated.
      Default constructor for the TwoStep class. This constructor initializes an instance of the TwoStep class without setting any specific properties or configurations.

  • Method Details

    • maskFromUndirected

      public static org.ejml.simple.SimpleMatrix maskFromUndirected(Graph skeleton, List<Node> vars)
      Deprecated.
      Constructs a mask matrix from an undirected graph, indicating the adjacency relationships between a given list of nodes. The resulting matrix is symmetric with off-diagonal elements set to 1.0 for adjacent nodes and 0.0 otherwise.
      Parameters:
      skeleton - the undirected graph representing the structure of the relationships.
      vars - the list of nodes for which the adjacency matrix is to be created.
      Returns:
      a symmetric mask matrix where an entry (i, j) is set to 1.0 if there is an adjacency between nodes vars[i] and vars[j], and 0.0 otherwise. The diagonal elements of the matrix are set to zero.

    • setLambda

      public void setLambda(double lambda)
      Deprecated.
      Sets the value of the regularization parameter lambda.
      Parameters:
      lambda - the new value for the regularization parameter

    • setNormalizeLossByN

      public void setNormalizeLossByN(boolean flag)
      Deprecated.
      If true, Step-1 uses lambdaEff = lambda * (ln N)/2 (paper-style).
      Parameters:
      flag - If normalized

    • setUseAdaptiveLasso

      public void setUseAdaptiveLasso(boolean flag)
      Deprecated.
      Sets whether the adaptive lasso method should be used in the two-step regression process.
      Parameters:
      flag - If true, enables the use of the adaptive lasso; otherwise, disables it.

    • setAdaptiveGamma

      public void setAdaptiveGamma(double gamma)
      Deprecated.
      Sets the value of the adaptive gamma parameter used in the two-step regression process.
      Parameters:
      gamma - the value to set for the adaptive gamma parameter

    • setMaskThreshold

      public void setMaskThreshold(double t)
      Deprecated.
      Sets the mask threshold value used in the two-step regression process.
      Parameters:
      t - the value to set for the mask threshold

    • setCoefThreshold

      public void setCoefThreshold(double t)
      Deprecated.
      Sets the coefficient threshold value for the regression process. This value affects regularization and the breaking of two-cycle constraints.
      Parameters:
      t - the new value to set for the coefficient threshold. It also ensures that the minimum absolute value for two-cycle checking is at least this threshold.

    • setIcaMaxIter

      public void setIcaMaxIter(int n)
      Deprecated.
      Sets the maximum number of iterations for the ICA (Independent Component Analysis) process. This parameter controls how many iterations the algorithm can perform before stopping.
      Parameters:
      n - the maximum number of iterations to allow for the ICA process

    • setIcaTol

      public void setIcaTol(double t)
      Deprecated.
      Sets the tolerance value for the ICA (Independent Component Analysis) process. This value determines the convergence criterion, defining the minimum change required between iterations for the algorithm to stop.
      Parameters:
      t - the tolerance value to set for the ICA process

    • setRandomSeed

      public void setRandomSeed(long seed)
      Deprecated.
      Sets the random seed for reproducibility in the algorithm's processes.
      Parameters:
      seed - the value of the random seed to set

    • setVerbose

      public void setVerbose(boolean v)
      Deprecated.
      Sets whether verbose output is enabled for the process.
      Parameters:
      v - true to enable verbose output; false to disable it.

    • setCondWarnThreshold

      public void setCondWarnThreshold(double thr)
      Deprecated.
      Sets the condition warning threshold value used in the processing. The threshold defines a boundary beyond which a condition is flagged for further attention.
      Parameters:
      thr - the threshold value to set for the condition warning

    • setFastIcaAlgorithm

      public void setFastIcaAlgorithm(int alg)
      Deprecated.
      Sets the FastICA algorithm type to be used in the Independent Component Analysis (ICA) procedure. This parameter determines the specific algorithmic approach to be applied within the FastICA process.
      Parameters:
      alg - the integer value representing the specific FastICA algorithm to use

    • setFastIcaFunction

      public void setFastIcaFunction(int f)
      Deprecated.
      Sets the FastICA function type to be used within the Independent Component Analysis (ICA) process. This function determines the contrast function applied during the execution of the FastICA algorithm.
      Parameters:
      f - the integer value representing the specific FastICA function to use

    • setFastIcaAlpha

      public void setFastIcaAlpha(double a)
      Deprecated.
      Sets the alpha parameter for the FastICA (Independent Component Analysis) process. This parameter is typically used to define the non-linearity or contrast function in the FastICA algorithm and influences its convergence and behavior.
      Parameters:
      a - the value of the FastICA alpha parameter to set

    • setFastIcaRowNorm

      public void setFastIcaRowNorm(boolean flag)
      Deprecated.
      Sets whether row normalization is applied in the FastICA (Independent Component Analysis) process. Enabling this parameter ensures that the rows of the data are normalized before the ICA computation, which can affect the outcome and convergence of the algorithm.
      Parameters:
      flag - true to enable row normalization in FastICA; false to disable it.

    • setBreakTwoCyclesEnabled

      public void setBreakTwoCyclesEnabled(boolean enabled)
      Deprecated.
      Enables or disables the breaking of two-cycle constraints in the processing.
      Parameters:
      enabled - true to enable the breaking of two-cycle constraints; false to disable it.

    • setTwoCycleMinAbs

      public void setTwoCycleMinAbs(double minAbs)
      Deprecated.
      Sets the minimum absolute value threshold for breaking two-cycle constraints in the processing. This value ensures that small values below the threshold are ignored when evaluating constraints.
      Parameters:
      minAbs - the threshold value to set for the minimum absolute value used in two-cycle checking

    • setExternalMask

      public void setExternalMask(org.ejml.simple.SimpleMatrix mask)
      Deprecated.
      Sets an external mask for use in the current object. The provided mask is sanitized before being assigned, and the external mask usage is enabled.
      Parameters:
      mask - the mask to be used as the external mask; it is represented as a SimpleMatrix object

    • getLastA

      public org.ejml.simple.SimpleMatrix getLastA()
      Deprecated.
      Retrieves the last calculated matrix A.
      Returns:
      the last instance of the SimpleMatrix referred to as lastA.

    • search

      public TwoStep.Result search(DataSet data)
      Deprecated.
      Performs a search operation over the given dataset to analyze dependencies, refine coefficients, and build a directed graph representation.

      This method utilizes techniques such as standardization, covariance analysis, masked projection, inverse estimation, constrained optimization, and additional cleanup steps to produce a refined result containing both a coefficient matrix and a graph representation.

      Parameters:
      data - The dataset to perform the search on. This should include observable variables and their associated data points.
      Returns:
      A Result object containing the refined coefficient matrix and the generated graph. The matrix represents the weighted relationships between variables, and the graph represents their directional dependencies.
      Throws:
      IllegalStateException - If the matrix inversion process fails or other critical computations encounter errors.