Update glyph net.

Author: ameritusweb

examples/gravnet/ParallelReverseAutoDiff.GravNetExample/GlyphNet.cs

@@ -80,7 +80,7 @@ public void AdjustLearningRate(double learningRate)
         /// <returns>The task.</returns>
         public async Task Initialize()
         {
-            var initialAdamIteration = 509;
+            var initialAdamIteration = 138;
             var model = new GlyphNetwork.GlyphNetwork(this.numLayers, this.numNodes, this.numFeatures, this.learningRate, this.clipValue, "glyphnet");
             model.Parameters.AdamIteration = initialAdamIteration;
             this.GlyphNetwork = model;
@@ -121,7 +121,7 @@ public void SaveWeights()
         /// </summary>
         public void ApplyWeights()
         {
-            var guid = "glyph_be1b3edd-d423-4714-ae20-3552bdc5cc9f_509";
+            var guid = "glyph_c83acea3-57af-4909-adab-0c75768549de_138";
             var dir = $"E:\\vnnstore\\{guid}";
             for (int i = 0; i < this.modelLayers.Count; ++i)
             {
@@ -160,16 +160,35 @@ public void ApplyGradients()
             var targetedSum0 = gatNet.TargetedSum0;
             var targetedSum1 = gatNet.TargetedSum1;
 
-            SquaredArclengthEuclideanLossOperation arclengthLoss0 = SquaredArclengthEuclideanLossOperation.Instantiate(gatNet);
-            var loss0 = arclengthLoss0.Forward(targetedSum0, (3 * Math.PI) / 4d);
+            int maxMag0 = 0;
+            int maxMag1 = 0;
+            for (int i = 0; i < 15; i++)
+            {
+                for (int j = 0; j < 15; j++)
+                {
+                    if (rotationTargets[i, j] == 1)
+                    {
+                        maxMag1++;
+                    }
+                    else
+                    {
+                        maxMag0++;
+                    }
+                }
+            }
+
+            Console.WriteLine($"Max Mag 0: {maxMag0}, Max Mag 1: {maxMag1}");
+
+            SquaredArclengthEuclideanMagnitudeLossOperation arclengthLoss0 = SquaredArclengthEuclideanMagnitudeLossOperation.Instantiate(gatNet);
+            var loss0 = arclengthLoss0.Forward(targetedSum0, (3 * Math.PI) / 4d, maxMag0);
             var gradient0 = arclengthLoss0.Backward();
 
-            SquaredArclengthEuclideanLossOperation arclengthLoss1 = SquaredArclengthEuclideanLossOperation.Instantiate(gatNet);
-            var loss1 = arclengthLoss1.Forward(targetedSum1, (1 * Math.PI) / 4d);
+            SquaredArclengthEuclideanMagnitudeLossOperation arclengthLoss1 = SquaredArclengthEuclideanMagnitudeLossOperation.Instantiate(gatNet);
+            var loss1 = arclengthLoss1.Forward(targetedSum1, (1 * Math.PI) / 4d, maxMag1);
             var gradient1 = arclengthLoss1.Backward();
 
-            SquaredArclengthEuclideanLossOperation arclengthLoss = SquaredArclengthEuclideanLossOperation.Instantiate(gatNet);
-            var loss = arclengthLoss.Forward(output, targetAngle);
+            SquaredArclengthEuclideanMagnitudeLossOperation arclengthLoss = SquaredArclengthEuclideanMagnitudeLossOperation.Instantiate(gatNet);
+            var loss = arclengthLoss.Forward(output, targetAngle, 225);
             var gradient = arclengthLoss.Backward();
 
             return (gradient, gradient0, gradient1, output, targetedSum0, targetedSum1, glyph, loss, loss0, loss1);

examples/gravnet/ParallelReverseAutoDiff.GravNetExample/GlyphNetTrainer.cs

@@ -14,11 +14,11 @@ public async Task Train()
                 CudaBlas.Instance.Initialize();
                 GlyphNet net = new GlyphNet(512, 6144, 3, 0.01d, 4d);
                 await net.Initialize();
-                //net.ApplyWeights();
+                net.ApplyWeights();
 
                 var pngFiles = Directory.GetFiles(@"E:\images\inputs\svg", "*.png");
 
-                Random random = new Random(7);
+                Random random = new Random(11);
                 var files = pngFiles.OrderBy(x => random.Next()).ToArray();
                 uint i = 0;
                 await files.WithRepeatAsync(async (pngFile, token) =>
@@ -93,14 +93,16 @@ await files.WithRepeatAsync(async (pngFile, token) =>
 
 
                     Console.WriteLine($"Iteration {i} Output X: {output[0, 0]}, Output Y: {output[0, 1]}, Grad: {gradient[0, 0]}, {gradient[0, 1]}");
-                    Console.WriteLine($"O1 X: {o1[0, 0]}, O1 Y: {o1[0, 1]}, Loss: {loss[0, 0]}, {loss0[0, 0]}, {loss1[0, 0]}");
+                    Console.WriteLine($"Loss: {loss[0, 0]}");
+                    Console.WriteLine($"O0 X: {o0[0, 0]}, O0 Y: {o0[0, 1]}, Loss 0: {loss0[0, 0]}");
+                    Console.WriteLine($"O1 X: {o1[0, 0]}, O1 Y: {o1[0, 1]}, Loss 1: {loss1[0, 0]}");
                     await net.Backward(gradient, gradient0, gradient1);
                     net.ApplyGradients();
                     //}
 
                     await net.Reset();
                     Thread.Sleep(1000);
-                    if (i % 11 == 2)
+                    if (i % 11 == 5)
                     {
                         net.SaveWeights();
                     }

examples/gravnet/ParallelReverseAutoDiff.GravNetExample/Program.cs

@@ -4,7 +4,7 @@ internal class Program
     {
         static void Main(string[] args)
         {
-            TiledNetTrainer trainer = new TiledNetTrainer();
+            GlyphNetTrainer trainer = new GlyphNetTrainer();
             trainer.Train().Wait();
         }
     }