edu.cmu.tetradapp.editor.datamanip.DeterminismEditor Maven / Gradle / Ivy
/*
* To change this license header, choose License Headers in Project Properties.
* To change this template file, choose Tools | Templates
* and open the template in the editor.
*/
package edu.cmu.tetradapp.editor.datamanip;
import edu.cmu.tetrad.data.*;
import edu.cmu.tetrad.graph.Node;
import edu.cmu.tetrad.graph.NodeVariableType;
import edu.cmu.tetrad.util.Parameters;
import edu.cmu.tetradapp.editor.FinalizingParameterEditor;
import edu.cmu.tetradapp.model.DataWrapper;
import org.apache.commons.lang3.SerializationUtils;
import javax.swing.*;
import javax.swing.table.DefaultTableModel;
import java.awt.*;
import java.util.List;
import java.util.*;
/**
* Feb 11, 2019 4:17:17 PM
*
* @author Zhou Yuan [email protected]
*/
public class DeterminismEditor extends JPanel implements FinalizingParameterEditor {
private static final long serialVersionUID = 6513664419620810219L;
private DataSet sourceDataSet;
private DataSet sourceDataSetCopy;
private DataSet mergedDataSet;
private Parameters parameters;
//==========================CONSTUCTORS===============================//
/**
* Constructs a new editor that will allow the user to merge determistic interventional variables. The editor will
* return the merged dataset
*/
public DeterminismEditor() {
}
//============================= Public Methods ===================================//
/**
* Sets up the GUI.
*/
@Override
public void setup() {
// Container
Box container = Box.createVerticalBox();
container.setPreferredSize(new Dimension(360, 200));
// Detect
List> mergedList = detectDeterministicVars();
// Intervention variables
Box mergedVarsBox = Box.createVerticalBox();
mergedVarsBox.setPreferredSize(new Dimension(320, 160));
mergedVarsBox.setBorder(BorderFactory.createTitledBorder("Deterministic Interventional Variables"));
// Merged variables table
JTable table = new JTable();
DefaultTableModel tableModel = new DefaultTableModel();
table.setModel(tableModel);
// Headers
List columnNames = Arrays.asList("Deterministic variables", "Merged new variable");
// Table header
tableModel.setColumnIdentifiers(columnNames.toArray());
mergedList.forEach(indexSet -> {
List varNameList = new LinkedList<>();
Integer[] indexArray = indexSet.toArray(new Integer[0]);
for (Integer integer : indexArray) {
varNameList.add(this.sourceDataSetCopy.getVariable(integer).getName());
}
String varNames = String.join(", ", varNameList);
String mergedVarName = String.join("_", varNameList);
List rowData = Arrays.asList(varNames, mergedVarName);
tableModel.addRow(rowData.toArray());
});
JScrollPane scrollPane = new JScrollPane(table);
mergedVarsBox.add(scrollPane);
// Add to container
container.add(mergedVarsBox);
// Adds the specified component to the end of this container.
add(container, BorderLayout.CENTER);
// Merge
mergeDeterministicVars(mergedList);
}
/**
* Tells the editor to commit any final details before it is closed (only called when the user selects "Ok" or
* something of that nature). If false is returned the edit is considered invalid and it will be treated as if the
* user selected "cancelAll".
*
* @return - true if the edit was committed.
*/
@Override
public boolean finalizeEdit() {
// This way we can pass the merged dataset to the wrapper
this.parameters.set("DeterminisedDataset", this.mergedDataSet);
return true;
}
/**
* Sets the previous params, must be DiscretizationParams.
*/
@Override
public void setParams(Parameters params) {
this.parameters = params;
}
/**
* The parent model should be a DataWrapper.
*/
@Override
public void setParentModels(Object[] parentModels) {
if (parentModels == null || parentModels.length == 0) {
throw new IllegalArgumentException("There must be parent model");
}
DataWrapper dataWrapper = null;
for (Object parent : parentModels) {
if (parent instanceof DataWrapper) {
dataWrapper = (DataWrapper) parent;
}
}
if (dataWrapper == null) {
throw new IllegalArgumentException("Should have have a data wrapper as a parent");
}
DataModel model = dataWrapper.getSelectedDataModel();
if (!(model instanceof DataSet)) {
throw new IllegalArgumentException("The dataset must be a rectangular dataset");
}
if (!containsInterventionalVariables(model)) {
throw new IllegalArgumentException("The dataset must contain interventional variables");
}
// Get the source dataset, keep it untouched
DataSet dataSet = (DataSet) model;
this.sourceDataSet = dataSet;
// Work on this dataset for merging, deep copy
this.sourceDataSetCopy = SerializationUtils.clone(dataSet);
}
@Override
public boolean mustBeShown() {
return true;
}
//=============================== Private Methods ================================//
/**
* Determinism detection among interventional variables
*/
private List> detectDeterministicVars() {
List> deterministicList = new ArrayList<>();
// Double for loop to build the initial deterministicList
for (int i = 0; i < this.sourceDataSetCopy.getVariables().size(); i++) {
Set set = new HashSet<>();
for (int j = i + 1; j < this.sourceDataSetCopy.getVariables().size(); j++) {
Node outerVar = this.sourceDataSetCopy.getVariable(i);
Node innerVar = this.sourceDataSetCopy.getVariable(j);
// Bidirectional check
if (isDeterministic(outerVar, innerVar) && isDeterministic(innerVar, outerVar)) {
if (outerVar.getNodeVariableType() == innerVar.getNodeVariableType()) {
set.add(j);
} else {
if ((outerVar.getNodeVariableType() != NodeVariableType.DOMAIN) && (innerVar.getNodeVariableType() != NodeVariableType.DOMAIN)) {
set.add(j);
} else {
// Add new attribute to domain variables that are fully determinised
// But don't merge
String fullyDeterminisedDomainVar = "fullyDeterminisedDomainVar";
if (outerVar.getNodeVariableType() == NodeVariableType.DOMAIN) {
outerVar.addAttribute(fullyDeterminisedDomainVar, true);
} else {
innerVar.addAttribute(fullyDeterminisedDomainVar, true);
}
}
}
}
}
// Add to list
deterministicList.add(set);
}
List> mergedList = new ArrayList<>();
for (int k = 0; k < deterministicList.size(); k++) {
// Create a new set for non-empty set and add all the elements of the sets whose index is in this set
if (!deterministicList.get(k).isEmpty()) {
Set mergedSet = new HashSet<>();
// Add the index of this set to the merged set
mergedSet.add(k);
mergedSet.addAll(deterministicList.get(k));
for (Integer index : deterministicList.get(k)) {
mergedSet.addAll(deterministicList.get(index));
// Then empty that set
deterministicList.get(index).clear();
}
// Finally add to the mergedList
mergedList.add(mergedSet);
}
}
return mergedList;
}
/**
* Merge the deterministic interventional variables and create new dataset
*/
private void mergeDeterministicVars(List> mergedList) {
List toBeRemovedColumns = new LinkedList<>();
mergedList.forEach(indexSet -> {
List varList = new LinkedList<>();
List varNameList = new LinkedList<>();
Integer[] indexArray = indexSet.toArray(new Integer[0]);
for (int i = 0; i < indexArray.length; i++) {
varList.add(this.sourceDataSetCopy.getVariable(indexArray[i]));
varNameList.add(this.sourceDataSetCopy.getVariable(indexArray[i]).getName());
// Add index for removal except the first one
// Remember the column to be removed after merging
if (i > 0) {
toBeRemovedColumns.add(indexArray[i]);
}
}
String mergedVarName = String.join("_", varNameList);
// Keep the first node as the merged node and rename
Node mergedNode = varList.get(0);
mergedNode.setName(mergedVarName);
//Use this hierarchy for determining the merged variable's type: [value] > [status]
NodeVariableType mergedNodeVariableType = mergedNode.getNodeVariableType();
for (Integer integer : indexArray) {
Node node = this.sourceDataSetCopy.getVariable(integer);
if ((node.getNodeVariableType() == NodeVariableType.INTERVENTION_VALUE) && (node.getNodeVariableType() != mergedNodeVariableType)) {
mergedNode.setNodeVariableType(NodeVariableType.INTERVENTION_VALUE);
break;
}
}
});
// Target variables
List nodeList = new LinkedList<>();
Map origIndexMap = new HashMap<>();
this.sourceDataSetCopy.getVariables().forEach(node -> {
int columnIndex = this.sourceDataSetCopy.getColumn(node);
// Skip these columns when creating the new dataset
if (!toBeRemovedColumns.contains(columnIndex)) {
nodeList.add(node);
origIndexMap.put(node, columnIndex);
}
});
// Now scan all the coloumns and create the data box
// Finally convert to data model
this.mergedDataSet = createDataBoxData(nodeList, origIndexMap);
}
/**
* Check to see if the dataset has interventional variables
*/
private boolean containsInterventionalVariables(DataModel model) {
List interventionalVars = new LinkedList<>();
model.getVariables().forEach(e -> {
if (e.getNodeVariableType() == NodeVariableType.INTERVENTION_STATUS || e.getNodeVariableType() == NodeVariableType.INTERVENTION_VALUE) {
interventionalVars.add(e.getName());
}
});
return interventionalVars.size() > 0;
}
/**
* Determine if discrete variable x and discrete variable y are deterministic
*/
private boolean isDeterministic(Node x, Node y) {
// For now only check between discrete variables
if ((x instanceof DiscreteVariable) && (y instanceof DiscreteVariable)) {
Map map = new HashMap<>();
int xColumnIndex = this.sourceDataSet.getColumn(x);
int yColumnIndex = this.sourceDataSet.getColumn(y);
int numRows = this.sourceDataSet.getNumRows();
for (int i = 0; i < numRows; i++) {
// objX as key, and objY as value
Object objX = this.sourceDataSet.getObject(i, xColumnIndex);
Object objY = this.sourceDataSet.getObject(i, yColumnIndex);
if (map.containsKey(objX)) {
if (!map.get(objX).equals(objY)) {
return false;
}
} else {
map.put(objX, objY);
}
}
return true;
}
return false;
}
/**
* Create BoxDataSet from the target nodes
*/
private BoxDataSet createDataBoxData(List nodeList, Map origIndexMap) {
// Now scan all the coloumns and create the data box
boolean isDiscrete = false;
boolean isContinuous = false;
for (Node node : nodeList) {
if (node instanceof ContinuousVariable) {
isContinuous = true;
} else {
isDiscrete = true;
}
}
if (isDiscrete && isContinuous) {
return createMixedDataBox(nodeList, origIndexMap);
} else if (isContinuous) {
return createContinuousDataBox(nodeList, origIndexMap);
} else if (isDiscrete) {
return createDiscreteDataBox(nodeList, origIndexMap);
} else {
return null;
}
}
private BoxDataSet createMixedDataBox(List nodeList, Map origIndexMap) {
int numOfCols = nodeList.size();
int numOfRows = this.sourceDataSetCopy.getNumRows();
double[][] continuousData = new double[numOfCols][];
int[][] discreteData = new int[numOfCols][];
for (int i = 0; i < numOfCols; i++) {
Node node = nodeList.get(i);
// initialize data
if (node instanceof DiscreteVariable) {
discreteData[i] = new int[numOfRows];
for (int j = 0; j < numOfRows; j++) {
discreteData[i][j] = this.sourceDataSetCopy.getInt(j, origIndexMap.get(node));
}
} else {
continuousData[i] = new double[numOfRows];
for (int j = 0; j < numOfRows; j++) {
continuousData[i][j] = this.sourceDataSetCopy.getDouble(j, origIndexMap.get(node));
}
}
}
return new BoxDataSet(new MixedDataBox(nodeList, numOfRows, continuousData, discreteData), nodeList);
}
/**
* The data can only be mixed or discrete. This won't be used for now
*/
private BoxDataSet createContinuousDataBox(List nodeList, Map origIndexMap) {
int numOfCols = nodeList.size();
int numOfRows = this.sourceDataSetCopy.getNumRows();
double[][] data = new double[numOfRows][numOfCols];
for (int i = 0; i < numOfRows; i++) {
for (int j = 0; j < numOfCols; j++) {
Node node = nodeList.get(j);
data[i][j] = this.sourceDataSetCopy.getDouble(i, origIndexMap.get(node));
}
}
return new BoxDataSet(new DoubleDataBox(data), nodeList);
}
private BoxDataSet createDiscreteDataBox(List nodeList, Map origIndexMap) {
int numOfCols = nodeList.size();
int numOfRows = this.sourceDataSetCopy.getNumRows();
int[][] data = new int[numOfCols][numOfRows];
for (int i = 0; i < numOfCols; i++) {
for (int j = 0; j < numOfRows; j++) {
Node node = nodeList.get(i);
data[i][j] = this.sourceDataSetCopy.getInt(j, origIndexMap.get(node));
}
}
return new BoxDataSet(new VerticalIntDataBox(data), nodeList);
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy