updated docs and package infos

MKLab-ITI · Aug 24, 2024 · e3cdc59 · e3cdc59
1 parent 899cd6d
commit e3cdc59
Show file tree

Hide file tree

Showing 383 changed files with 8,793 additions and 5,394 deletions.
diff --git a/JGNN/src/examples/classification/LogisticRegression.java b/JGNN/src/examples/classification/LogisticRegression.java
@@ -2,10 +2,10 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.nn.loss.Accuracy;
 import mklab.JGNN.nn.loss.BinaryCrossEntropy;
 import mklab.JGNN.core.Slice;

diff --git a/JGNN/src/examples/classification/MLP.java b/JGNN/src/examples/classification/MLP.java
@@ -2,10 +2,10 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/nodeClassification/APPNP.java b/JGNN/src/examples/nodeClassification/APPNP.java
@@ -3,11 +3,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/nodeClassification/GAT.java b/JGNN/src/examples/nodeClassification/GAT.java
@@ -2,11 +2,12 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
+import mklab.JGNN.adhoc.train.NodeClassification;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;
@@ -34,7 +35,7 @@ public static void main(String[] args) throws Exception {
 				.classify()
 				.assertBackwardValidity();
 
-		ModelTraining trainer = new ModelTraining()
+		ModelTraining trainer = new NodeClassification()
 				.setOptimizer(new Adam(0.01))
 				.setEpochs(300)
 				.setPatience(100)
@@ -48,7 +49,9 @@ public static void main(String[] args) throws Exception {
 				.init(new XavierNormal())
 				.train(trainer,
 						Tensor.fromRange(nodes.size()).asColumn(), 
-						dataset.labels(), nodes.range(0, 0.6), nodes.range(0.6, 0.8));
+						dataset.labels(), 
+						nodes.range(0, 0.6), 
+						nodes.range(0.6, 0.8));
 
 		System.out.println("Training time "+(System.currentTimeMillis()-tic)/1000.);
 		Matrix output = model.predict(Tensor.fromRange(0, nodes.size()).asColumn()).get(0).cast(Matrix.class);

diff --git a/JGNN/src/examples/nodeClassification/GCN.java b/JGNN/src/examples/nodeClassification/GCN.java
@@ -5,11 +5,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.core.empy.EmptyTensor;

diff --git a/JGNN/src/examples/nodeClassification/GCNII.java b/JGNN/src/examples/nodeClassification/GCNII.java
@@ -2,11 +2,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/nodeClassification/HetGCN.java b/JGNN/src/examples/nodeClassification/HetGCN.java
@@ -2,11 +2,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/nodeClassification/MessagePassing.java b/JGNN/src/examples/nodeClassification/MessagePassing.java
@@ -2,11 +2,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/nodeClassification/Scripting.java b/JGNN/src/examples/nodeClassification/Scripting.java
@@ -4,11 +4,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.Neuralang;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.core.empy.EmptyTensor;

diff --git a/JGNN/src/examples/tutorial/Learning.java b/JGNN/src/examples/tutorial/Learning.java
@@ -2,11 +2,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/tutorial/NN.java b/JGNN/src/examples/tutorial/NN.java
@@ -3,11 +3,11 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Citeseer;
 import mklab.JGNN.adhoc.parsers.LayeredBuilder;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.nn.initializers.XavierNormal;

diff --git a/JGNN/src/examples/tutorial/Quickstart.java b/JGNN/src/examples/tutorial/Quickstart.java
@@ -2,11 +2,12 @@
 
 import mklab.JGNN.adhoc.Dataset;
 import mklab.JGNN.adhoc.ModelBuilder;
+import mklab.JGNN.adhoc.ModelTraining;
 import mklab.JGNN.adhoc.datasets.Cora;
 import mklab.JGNN.adhoc.parsers.FastBuilder;
+import mklab.JGNN.adhoc.parsers.Neuralang;
 import mklab.JGNN.core.Matrix;
 import mklab.JGNN.nn.Model;
-import mklab.JGNN.nn.ModelTraining;
 import mklab.JGNN.core.Slice;
 import mklab.JGNN.core.Tensor;
 import mklab.JGNN.core.empy.EmptyTensor;
@@ -25,42 +26,49 @@ public class Quickstart {
 	public static void main(String[] args) throws Exception {
 		Dataset dataset = new Cora();
 		dataset.graph().setMainDiagonal(1).setToSymmetricNormalization();
-
+
+		String architectures = """
+			fn classify(nodes, h, epochs: !3000, patience: !100, lr: !0.01) {
+				return softmax(h[nodes], dim: "row");
+			}
+			fn gcnlayer(A, h, hidden: 16, reg: 0.005) {
+				return A@h@matrix(?, hidden, reg) + vector(hidden);
+			}
+			fn gcn(A, h, classes: extern) {
+				h = gcnlayer(A, h);
+				h = dropout(relu(h), 0.5);
+				return gcnlayer(A, h, hidden: classes);
+			}
+		""";
+
 		long numSamples = dataset.samples().getSlice().size();
 		long numClasses = dataset.labels().getCols();
-		ModelBuilder modelBuilder = new FastBuilder(dataset.graph(), dataset.features())
-			.config("reg", 0.005)
-			.config("classes", numClasses)
-			.config("hidden", 64)
-			.function("gcnlayer", "(A,h){Adrop = dropout(A, 0.5); return Adrop@(h@matrix(?, hidden, reg))+vector(?);}")
-			.layer("h{l+1}=relu(gcnlayer(A, h{l}))")
-			.config("hidden", "classes")  // reassigns the output gcnlayer's "hidden" to be the number of "classes"
-			.layer("h{l+1}=gcnlayer(A, h{l})")
-			.classify()
-			.autosize(new EmptyTensor(numSamples));
+		ModelBuilder modelBuilder = new Neuralang()
+				.parse(architectures)
+				.constant("A", dataset.graph())
+				.constant("h", dataset.features())
+				.var("nodes")
+				.config("classes", numClasses)
+				.config("hidden", numClasses+2)
+				.out("classify(nodes, gcn(A,h))")
+				.autosize(new EmptyTensor(numSamples));
+		System.out.println("Preferred learning rate: "+modelBuilder.getConfig("lr"));
 
 		ModelTraining trainer = new ModelTraining()
-				.setOptimizer(new Adam(0.01))
-				.setEpochs(3000)
-				.setPatience(100)
+				.configFrom(modelBuilder)
 				.setLoss(new CategoricalCrossEntropy())
-				.setValidationLoss(new VerboseLoss(new Accuracy()).setInterval(10));
+				.setValidationLoss(new VerboseLoss(new CategoricalCrossEntropy()));
 
-		Slice nodes = dataset.samples().getSlice().shuffle(); // a permutation of node identifiers
-		Matrix inputData = Tensor.fromRange(nodes.size()).asColumn(); // each node has its identifier as an input
+		long tic = System.currentTimeMillis();
+		Slice nodes = dataset.samples().getSlice().shuffle(100);
 		Model model = modelBuilder.getModel()
 				.init(new XavierNormal())
-				.train(trainer, 
-						inputData,
-						dataset.labels(), 
-						nodes.range(0, 0.6),  // train slice
-						nodes.range(0.6, 0.8)  // validation slice
-						);
-
-		//modelBuilder.save(Paths.get("gcn_cora.jgnn"));
+				.train(trainer,
+						Tensor.fromRange(nodes.size()).asColumn(), 
+						dataset.labels(), nodes.range(0, 0.6), nodes.range(0.6, 0.8));
 
-		Model loadedModel = model;//ModelBuilder.load(Paths.get("gcn_cora.jgnn")).getModel();
-		Matrix output = loadedModel.predict(Tensor.fromRange(0, nodes.size()).asColumn()).get(0).cast(Matrix.class);
+		System.out.println("Training time "+(System.currentTimeMillis()-tic)/1000.);
+		Matrix output = model.predict(Tensor.fromRange(0, nodes.size()).asColumn()).get(0).cast(Matrix.class);
 		double acc = 0;
 		for(Long node : nodes.range(0.8, 1)) {
 			Matrix nodeLabels = dataset.labels().accessRow(node).asRow();