Merge pull request #78 from IDSIA/dev

Release 0.2.0

Author: cbonesana

CHANGELOG.md

@@ -1,3 +1,19 @@
+0.2.0
+===========
+
+- New Factor hierarchy ( #74 ). Major changes are:
+  * All factors are now **interfaces**.
+  * For each `factor` we can have multiple implementations.
+  * `Factor`s are **immutable** now: the creation of a new factor is delegated to the `constructor` or to a dedicated `Factory` class.
+  * `SymbolicFactor`s can now be used to keep track of operations done by algorithms.
+- Added code for isipta21benchamrk ( #76 #77 )
+- Removed old methods marked as `@deprecated`
+
+0.1.7.a
+===========
+
+- Fixed an issue with sampling of `BayesianFactors` #75
+
 0.1.7
 ===========
 

README.md

@@ -10,45 +10,49 @@ learning and inference algorithms for credal models.
 An example of exact inference in a credal network is given below.
 
 ```java
-import ch.idsia.crema.factor.credal.vertex.VertexFactor;
-import ch.idsia.crema.inference.ve.CredalVariableElimination;
 import ch.idsia.crema.core.ObservationBuilder;
 import ch.idsia.crema.core.Strides;
+import ch.idsia.crema.factor.credal.vertex.separate.VertexFactor;
+import ch.idsia.crema.factor.credal.vertex.separate.VertexFactorFactory;
+import ch.idsia.crema.inference.ve.CredalVariableElimination;
+import ch.idsia.crema.model.graphical.DAGModel;
 import ch.idsia.crema.model.graphical.GraphicalModel;
 
 public class Starting {
-    public static void main(String[] args) {
-        double p = 0.2;
-        double eps = 0.0001;
-
-        /*  CN defined with vertex Factor  */
-
-        // Define the model (with vertex factors)
-        GraphicalModel<VertexFactor> model = new DAGModel<>();
-              int A = model.addVariable(3);
-        int B = model.addVariable(2);
-        
-        model.addParent(B,A);
-
-        // Define a credal set of the partent node
-        VertexFactor fu = new VertexFactor(model.getDomain(A), Strides.empty());
-        fu.addVertex(new double[]{0., 1-p, p});
-        fu.addVertex(new double[]{1-p, 0., p});
-        
-        model.setFactor(A,fu);
-
-        // Define the credal set of the child
-        VertexFactor fx = new VertexFactor(model.getDomain(B), model.getDomain(A));
-        fx.addVertex(new double[]{1., 0.,}, 0);
-        fx.addVertex(new double[]{1., 0.,}, 1);
-        fx.addVertex(new double[]{0., 1.,}, 2);
-
-        model.setFactor(B,fx);
-
-        // Run exact inference
-        CredalVariableElimination<VertexFactor> inf = new CredalVariableElimination<>(model);
-        inf.query(A, ObservationBuilder.observe(B,0));
-    }
+	public static void main(String[] args) {
+		double p = 0.2;
+		double eps = 0.0001;
+
+		/*  CN defined with vertex Factor  */
+
+		// Define the model (with vertex factors)
+		GraphicalModel<VertexFactor> model = new DAGModel<>();
+		int A = model.addVariable(3);
+		int B = model.addVariable(2);
+
+		model.addParent(B, A);
+
+		// Define a credal set of the partent node
+		VertexFactor fu = VertexFactorFactory.factory().domain(model.getDomain(A), Strides.empty())
+				.addVertex(new double[]{0., 1 - p, p})
+				.addVertex(new double[]{1 - p, 0., p})
+				.get();
+
+		model.setFactor(A, fu);
+
+		// Define the credal set of the child
+		VertexFactor fx = VertexFactorFactory.factory().domain(model.getDomain(B), model.getDomain(A))
+				.addVertex(new double[]{1., 0.,}, 0)
+				.addVertex(new double[]{1., 0.,}, 1)
+				.addVertex(new double[]{0., 1.,}, 2)
+				.get();
+
+		model.setFactor(B, fx);
+
+		// Run exact inference
+		CredalVariableElimination inf = new CredalVariableElimination();
+		inf.query(model, ObservationBuilder.observe(B, 0), A);
+	}
 }
 ``` 
 
@@ -68,7 +72,7 @@ Add the following code in the  pom.xml of your project:
         <dependency>
             <groupId>ch.idsia</groupId>
             <artifactId>crema</artifactId>
-            <version>0.1.6</version>
+            <version>0.1.7</version>
             <scope>compile</scope>
         </dependency>
     </dependencies>

examples/BNdefinition.java

@@ -1,52 +1,51 @@
+package examples;
+
+import ch.idsia.crema.factor.bayesian.BayesianDefaultFactor;
 import ch.idsia.crema.factor.bayesian.BayesianFactor;
 import ch.idsia.crema.model.graphical.BayesianNetwork;
 
 
 public class BNdefinition {
-    public static void main(String[] args) {
-
-        BayesianNetwork model = new BayesianNetwork();
-        BayesianFactor[] f = new BayesianFactor[5];
-
-        // Winter?
-        int A = model.addVariable(2);
-        f[A] = new BayesianFactor(model.getDomain(A), new double[]{.6, .4}, false);
-
-        // Sprinkler?
-        int B = model.addVariable(2);
-        model.addParent(B, A);
-        f[B] = new BayesianFactor(model.getDomain(A, B), false);
-        f[B].setData(new int[]{B, A}, new double[]{.2, .8, .75, .25});
-
-        // Rain?
-        int C = model.addVariable(2);
-        model.addParent(C, A);
-        f[C] = new BayesianFactor(model.getDomain(A, C), false);
-        f[C].setData(new int[]{C, A}, new double[]{.8, .2, .1, .9});
-
-        // Wet Grass?
-        int D = model.addVariable(2);
-        model.addParent(D, B);
-        model.addParent(D, C);
-        f[D] = new BayesianFactor(model.getDomain(B, C, D), false);
-        f[D].setData(new int[]{D, B, C}, new double[]{.95, .05, .9, .1, .8, .2, 0, 1});
-
-        // Slippery Road?
-        int E = model.addVariable(2);
-        model.addParent(E, C);
-        f[E] = new BayesianFactor(model.getDomain(C, E), false);
-        f[E].setData(new int[]{E, C}, new double[]{.7, .3, 0, 1});
-
-        model.setFactors(f);
-
-        System.out.println(model);
-
-        for(int x: model.getVariables()){
-            for(int y: model.getChildren(x)){
-                System.out.print("("+x+"-->"+y+")");
-            }
-        }
-
-    }
+	public static void main(String[] args) {
+
+		BayesianNetwork model = new BayesianNetwork();
+		BayesianFactor[] f = new BayesianFactor[5];
+
+		// Winter?
+		int A = model.addVariable(2);
+		f[A] = new BayesianDefaultFactor(model.getDomain(A), new double[]{.6, .4});
+
+		// Sprinkler?
+		int B = model.addVariable(2);
+		model.addParent(B, A);
+		f[B] = new BayesianDefaultFactor(model.getDomain(A, B), new int[]{B, A}, new double[]{.2, .8, .75, .25});
+
+		// Rain?
+		int C = model.addVariable(2);
+		model.addParent(C, A);
+		f[C] = new BayesianDefaultFactor(model.getDomain(A, C), new int[]{C, A}, new double[]{.8, .2, .1, .9});
+
+		// Wet Grass?
+		int D = model.addVariable(2);
+		model.addParent(D, B);
+		model.addParent(D, C);
+		f[D] = new BayesianDefaultFactor(model.getDomain(B, C, D), new int[]{D, B, C}, new double[]{.95, .05, .9, .1, .8, .2, 0, 1});
+
+		// Slippery Road?
+		int E = model.addVariable(2);
+		model.addParent(E, C);
+		f[E] = new BayesianDefaultFactor(model.getDomain(C, E), new int[]{E, C}, new double[]{.7, .3, 0, 1});
+
+		model.setFactors(f);
+
+		System.out.println(model);
+
+		for (int x : model.getVariables()) {
+			for (int y : model.getChildren(x)) {
+				System.out.print("(" + x + "-->" + y + ")");
+			}
+		}
+
+	}
 }
 //54

examples/BayesianFactors.java

@@ -1,53 +1,61 @@
+package examples;
+
 import ch.idsia.crema.core.Domain;
 import ch.idsia.crema.core.DomainBuilder;
+import ch.idsia.crema.factor.bayesian.BayesianDefaultFactor;
 import ch.idsia.crema.factor.bayesian.BayesianFactor;
+import ch.idsia.crema.factor.bayesian.BayesianFactorFactory;
 import ch.idsia.crema.model.graphical.BayesianNetwork;
 
 
 public class BayesianFactors {
-    public static void main(String[] args) {
-
-        // Create domain of 2 variables of sizes 2,3
-        Domain d1 = DomainBuilder.var(0, 1).size(2, 3);
+	public static void main(String[] args) {
 
-        // Some operations over a domain
-        d1.getVariables();              // vector of variables (int[])
-        d1.getSize();                   // number of variables
-        d1.getCardinality(0);    // size of a specific variable
-        d1.getSizeAt(0);
+		// Create domain of 2 variables of sizes 2,3
+		Domain d1 = DomainBuilder.var(0, 1).size(2, 3);
 
+		// Some operations over a domain
+		d1.getVariables();              // vector of variables (int[])
+		d1.getSize();                   // number of variables
+		d1.getCardinality(0);    // size of a specific variable
+		d1.getSizeAt(0);
 
-        // Crate a factor over the domain
-        BayesianFactor f = new BayesianFactor(d1);  // P([0, 1])  ->  P([X|Y])
 
-        f.getDomain();
+		// Crate a factor over the domain and set the data (i.e., values) of the factor
+		BayesianFactor f = new BayesianDefaultFactor(d1,  // P([0, 1])  ->  P([X|Y])
+				new double[]{
+						// x1 x2
+						0.2, 0.8, // y1
+						0.5, 0.5, // y2
+						0.1, 0.9  // y3
+				});
+		f.getDomain();
 
-        // Set the data (i.e., values) of the factor
-        //                      x1   x2
-        f.setData(new double[]{0.2, 0.8,   //y1
-                0.5, 0.5,   //y2
-                0.1, 0.9    //y3
-        });
+		// or using the dedicated Factory class:
+		f = BayesianFactorFactory.factory()
+				.domain(d1)
+				.data(new double[]{0.2, 0.8, 0.5, 0.5, 0.1, 0.9})
+				.get();
 
-        f.marginalize(0).getData(); // = double[3] { 1.0, 1.0, 1.0 }
+		f.marginalize(0).getData(); // = double[3] { 1.0, 1.0, 1.0 }
 
 
-        ////// A factor can also be defined in the context of a specific model
+		////// A factor can also be defined in the context of a specific model
 
 
-        // Define the BN
-        BayesianNetwork model = new BayesianNetwork();
-        model.addVariables(2, 3);
-        model.addParent(0, 1);
+		// Define the BN
+		BayesianNetwork model = new BayesianNetwork();
+		model.addVariables(2, 3);
+		model.addParent(0, 1);
 
-        // extract a domain with varaiables 0 and 1 in the BN
-        Domain d2 = model.getDomain(0, 1);
+		// extract a domain with varaiables 0 and 1 in the BN
+		Domain d2 = model.getDomain(0, 1);
 
-        // Create an empty factor with that domain
-        f = new BayesianFactor(model.getDomain(0, 1));
+		// Create an empty factor with that domain
+		f = new BayesianDefaultFactor(model.getDomain(0, 1), null);
 
 
-    }
+	}
 }
 //53