Package edu.cmu.tetrad.search.unmix

Class EmUnmix.Config

java.lang.Object

edu.cmu.tetrad.search.unmix.EmUnmix.Config

Enclosing class:

EmUnmix

public static final class EmUnmix.Config
extends Object

Configuration class for the EmUnmix algorithm, providing parameters and settings to control the behavior of the unmixing process. This class encapsulates various options for clustering, initialization, and residual scaling. The properties allow users to configure aspects such as the number of clusters, EM algorithm behavior, covariance type, regularization, randomization, annealing, and other advanced settings.

Field Summary

Fields

Modifier and Type	Field	Description
double	annealStartT	Represents the starting temperature for the simulated annealing algorithm.
int	annealSteps	Represents the random seed for initializing the random number generator.
double	covRidgeRel	Represents the annealing steps for the simulated annealing algorithm.
double	covShrinkage	Represents the starting temperature for the simulated annealing algorithm.
GaussianMixtureEM.CovarianceType	covType	Specifies the covariance type to be used in the EmUnmix algorithm's Gaussian Mixture Model.
int	emMaxIters	Maximum number of iterations allowed for the Expectation-Maximization (EM) algorithm during the computation process.
double	emTol	Convergence tolerance for the Expectation-Maximization (EM) algorithm.
int	K	The number of clusters to be used in the EmUnmix algorithm.
int	kmeansRestarts	Represents the number of restarts for the K-means clustering algorithm.
long	randomSeed	Represents the seed value used to initialize a random number generator.
double	ridge	Represents the ridge parameter for regularization in the covariance matrix estimation.
boolean	robustScaleResiduals	Determines whether robust scaling should be applied to the residuals during the EmUnmix algorithm process.
long	seed	Represents the seed value used for initializing random number generation or other operations that require a deterministic starting point.
ParentSupersetBuilder.Config	supersetCfg	Configuration object for the parent superset used in the EmUnmix algorithm.
boolean	useMAP	Indicates whether to use Maximum A Posteriori (MAP) estimation for parameter initialization.
boolean	useParentSuperset	Indicates whether to use a parent superset for subset initialization in the EmUnmix algorithm.

Constructor Summary

Constructors

Constructor	Description
Config()	Default constructor for Config.

Method Summary

Modifier and Type	Method	Description
EmUnmix.Config	copy()	Creates a copy of this configuration object.

Methods inherited from class java.lang.Object

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

Field Details

K

public int K

The number of clusters to be used in the EmUnmix algorithm. This defines the number of distinct components in the Gaussian Mixture Model, influencing the clustering and unmixing process. A higher value of K allows for more clusters, which may better capture the data structure but could also lead to overfitting.

useParentSuperset

public boolean useParentSuperset

Indicates whether to use a parent superset for subset initialization in the EmUnmix algorithm. This flag controls whether the algorithm should leverage a pre-defined superset configuration as a starting point for cluster construction. When set to true, it enables the algorithm to utilize a higher-level grouping structure to aid in the initialization process, potentially improving clustering results in scenarios with hierarchical data organization.

supersetCfg

public ParentSupersetBuilder.Config supersetCfg

Configuration object for the parent superset used in the EmUnmix algorithm. This variable encapsulates settings specific to the initialization and construction of cluster supersets. It provides configurable parameters for guiding the unmixing process by utilizing a parent superset structure.

robustScaleResiduals

public boolean robustScaleResiduals

Determines whether robust scaling should be applied to the residuals during the EmUnmix algorithm process. When set to true, the algorithm applies robust techniques to normalize the residuals, reducing the influence of outliers and improving the stability and reliability of the unmixing process.

covType

public GaussianMixtureEM.CovarianceType covType

Specifies the covariance type to be used in the EmUnmix algorithm's Gaussian Mixture Model. This variable determines how the covariance matrices are modeled for the different Gaussian components in the mixture. The choice of covariance type can influence the flexibility and complexity of the model when fitting the data.

Possible values include:

• FULL: Allows for a full covariance matrix, capturing correlations between all variables.
• DIAGONAL: Restricts the covariance matrix to be diagonal, assuming no correlations between variables.
• SPHERICAL: Assumes equal variance (spherical covariance) for all dimensions.

emMaxIters

public int emMaxIters

Maximum number of iterations allowed for the Expectation-Maximization (EM) algorithm during the computation process. This value places an upper limit on the iterations to ensure termination and prevent excessive computation time.

emTol

public double emTol

Convergence tolerance for the Expectation-Maximization (EM) algorithm. This parameter specifies the threshold for detecting convergence in the iterative optimization process of the EM algorithm. Smaller values indicate stricter convergence criteria, while larger values may result in faster but less precise convergence.

seed

public long seed

Represents the seed value used for initializing random number generation or other operations that require a deterministic starting point. This value ensures reproducibility of results when running stochastic processes or algorithms that involve randomness.

ridge

public double ridge

Represents the ridge parameter for regularization in the covariance matrix estimation. A positive value adds a penalty term to the covariance matrix to prevent overfitting.

kmeansRestarts

public int kmeansRestarts

Represents the number of restarts for the K-means clustering algorithm. Increasing this value can improve the quality of the clustering but may also increase computation time.

useMAP

public boolean useMAP

Indicates whether to use Maximum A Posteriori (MAP) estimation for parameter initialization. MAP estimation can provide more stable and accurate parameter estimates.

covRidgeRel

public double covRidgeRel

Represents the annealing steps for the simulated annealing algorithm. Increasing this value can improve the quality of the solution but may also increase computation time.

covShrinkage

public double covShrinkage

Represents the starting temperature for the simulated annealing algorithm. A higher value can lead to more exploration of the solution space.

annealSteps

public int annealSteps

Represents the random seed for initializing the random number generator. This ensures reproducibility of results when running stochastic algorithms.

annealStartT

public double annealStartT

Represents the starting temperature for the simulated annealing algorithm. A higher value can lead to more exploration of the solution space.

randomSeed

public long randomSeed

Represents the seed value used to initialize a random number generator. This value ensures that results are reproducible by producing a consistent sequence of random numbers when the same seed is used.

Constructor Details

Config

public Config()

Default constructor for Config.

Method Details

copy

public EmUnmix.Config copy()

Creates a copy of this configuration object.

Returns:

a new Config object with the same settings as this one.