Added generator and discriminator network balancing (disabled for now)

And other changes

Author: generic-github-user

script.js

@@ -7,7 +7,7 @@ const numParameters = 4;
 const imageSize = 8;
 // Number of images to use when training the neural network
 const numTrainingImages = 1;
-const logData = false;
+const logData = true;
 const optimizer = {
 "generator": tf.train.adam(0.001),
 "discriminator": tf.train.sgd(0.001)
@@ -223,8 +223,8 @@ trainingData.images[trainingData.images.length - 1].onload = function () {
 generateTrainingData();
 }
 
-generatorLoss = generator.calculateLoss();
-discriminatorLoss = discriminator.calculateLoss();
+const generatorLoss = generator.calculateLoss();
+const discriminatorLoss = discriminator.calculateLoss();
 
 if (logData) {
 console.log("Iteration " + iteration);
@@ -245,16 +245,11 @@ trainingData.images[trainingData.images.length - 1].onload = function () {
 }
 document.querySelector("#iteration").innerHTML = "Iteration • " + iteration;
 document.querySelector("#generator-loss").innerHTML = "Generator • " +
-generatorLoss.
-dataSync()[0].
-toFixed(2);
+generatorLoss
+.dataSync()[0];
 document.querySelector("#discriminator-loss").innerHTML = "Discriminator • " +
-discriminatorLoss.
-dataSync()[0].
-toFixed(2);
-
-generatorLoss.dispose();
-discriminatorLoss.dispose();
+discriminatorLoss
+.dataSync()[0];
 
 const trainableVars = [];
 for (var i = 0; i < generator.model.weights.length; i ++) {
@@ -263,12 +258,24 @@ trainingData.images[trainingData.images.length - 1].onload = function () {
 for (var i = 0; i < generator.model.model.weights.length; i ++) {
 trainableVars.push(generator.model.model.weights[i].val);
 }
+// if (generatorLoss > discriminatorLoss) {
 generator.optimizer.minimize(
 generator.calculateLoss,
 false,
 trainableVars
 );
 // }
+// else {
+discriminator.optimizer.minimize(discriminator.calculateLoss);
+// }
+
+// if (discriminatorLoss < 0.05) {
+//	generator.optimizer.minimize(generator.calculateLoss);
+// }
+// discriminator.optimizer.minimize(discriminator.calculateLoss);
+
+generatorLoss.dispose();
+discriminatorLoss.dispose();
 
 // All this is just display code
 // Calculate autoencoder output from original image
@@ -280,6 +287,7 @@ trainingData.images[trainingData.images.length - 1].onload = function () {
 return generator.model.predict(parameters.display)
 // Clip pixel values to a 0 - 255 (int32) range
 // Reshape the output tensor into an image format (W * L * 3)
+.clipByValue(0, 255)
 .reshape(
 [imageSize, imageSize, 1]
 )
@@ -291,21 +299,21 @@ trainingData.images[trainingData.images.length - 1].onload = function () {
 tf.tidy(
 () => {
 return discriminator.model.predict(
-generator.model.predict(parameters.display).clipByValue(0, 255)
+generator.model.predict(parameters.display)
 );
 }
 );
 
 if (logData) {
 console.log("Generator output");
-output.print();
+// output.print();
 
 console.log("Discriminator output");
 discriminatorOutput.print();
 }
 
 // Display the output tensor on the output canvas, then dispose the tensor
-tf.toPixels(output.clipByValue(0, 255), canvas.generated).then(() => output.dispose());
+tf.toPixels(output, canvas.generated).then(() => output.dispose());
 discriminatorOutput.dispose();
 
 iteration ++;