SameDiff loss functions (#6534)

* First pass on properly blocking out loss functions * Cleanup and loss function methods * First steps for new loss tests * Build out loss tests; Huber op arg fix * Javadoc, test fixes * More test fixes, javadoc pass * Clean up old code; more tests * Weighted loss function tests; ignore case with logged issue * Fix MPWSE loss function (no reduce mode as not supported) * Improvements to weighted loss tests * Javadoc and cleanup
deeplearning4j · Oct 5, 2018 · 409a706 · 409a706
1 parent 6be8fd0
commit 409a706
Show file tree

Hide file tree

Showing 20 changed files with 791 additions and 1,218 deletions.
diff --git a/...arent/nd4j-api/src/main/java/org/nd4j/autodiff/functions/DifferentialFunctionFactory.java b/...arent/nd4j-api/src/main/java/org/nd4j/autodiff/functions/DifferentialFunctionFactory.java
@@ -20,6 +20,7 @@
 import lombok.NonNull;
 import lombok.val;
 import org.apache.commons.lang3.ArrayUtils;
+import org.nd4j.autodiff.loss.LossReduce;
 import org.nd4j.autodiff.samediff.SDVariable;
 import org.nd4j.autodiff.samediff.SameDiff;
 import org.nd4j.base.Preconditions;
@@ -36,6 +37,7 @@
 import org.nd4j.linalg.api.ops.impl.indexaccum.*;
 import org.nd4j.linalg.api.ops.impl.layers.convolution.*;
 import org.nd4j.linalg.api.ops.impl.layers.convolution.config.*;
+import org.nd4j.linalg.api.ops.impl.loss.*;
 import org.nd4j.linalg.api.ops.impl.scalar.*;
 import org.nd4j.linalg.api.ops.impl.scalar.comparison.*;
 import org.nd4j.linalg.api.ops.impl.scatter.*;
@@ -1308,101 +1310,43 @@ public SDVariable weightedCrossEntropyWithLogits(SDVariable targets, SDVariable
         return new WeightedCrossEntropyLoss(sameDiff(), targets, inputs, weights).outputVariable();
     }
 
-    public SDVariable sigmoidCrossEntropyWithLogits(SDVariable logits, SDVariable weights, SDVariable labels,
-                                                    int reductionMode, double labelSmoothing) {
-        return new SigmoidCrossEntropyLoss(sameDiff(), logits, weights, labels,
-                reductionMode, labelSmoothing).outputVariable();
+    public SDVariable lossAbsoluteDifference(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce){
+        return new AbsoluteDifferenceLoss(sameDiff(), lossReduce, predictions, weights, label).outputVariable();
     }
 
-    public SDVariable softmaxCrossEntropyWithLogits(SDVariable logits, SDVariable weights, SDVariable labels,
-                                                    int reductionMode, double labelSmoothing) {
-        return new SoftmaxCrossEntropyLoss(sameDiff(), logits, weights, labels,
-                reductionMode, labelSmoothing).outputVariable();
+    public SDVariable lossCosineDistance(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce, int dimension){
+        return new CosineDistanceLoss(sameDiff(), lossReduce, predictions, weights, label, dimension).outputVariable();
     }
 
-    public SDVariable lossBinaryXENT(SDVariable iX,
-                                     SDVariable i_y,
-                                     int... dimensions) {
-        throw new UnsupportedOperationException();
+    public SDVariable lossHinge(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce){
+        return new HingeLoss(sameDiff(), lossReduce, predictions, weights, label).outputVariable();
     }
 
-
-    public SDVariable lossCosineSimilarity(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-    }
-
-
-    public SDVariable lossHinge(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
-
-    public SDVariable lossKLD(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
-
-    public SDVariable lossL1(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
-
-    public SDVariable lossL2(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
-
-    public SDVariable lossMAE(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossHuber(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce, double delta){
+        return new HuberLoss(sameDiff(), lossReduce, predictions, weights, label, delta).outputVariable();
     }
 
-
-    public SDVariable lossMAPE(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossLog(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce, double epsilon){
+        return new LogLoss(sameDiff(), lossReduce, predictions, weights, label, epsilon).outputVariable();
     }
 
-
-    public SDVariable lossMSE(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossMeanPairwiseSquaredError(SDVariable label, SDVariable predictions, SDVariable weights){
+        return new MeanPairwiseSquaredErrorLoss(sameDiff(), predictions, weights, label).outputVariable();
     }
 
-
-    public SDVariable lossMCXENT(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossMeanSquaredError(SDVariable label, SDVariable predictions, SDVariable weights, LossReduce lossReduce){
+        return new MeanSquaredErrorLoss(sameDiff(), lossReduce, predictions, weights, label).outputVariable();
     }
 
-
-    public SDVariable lossMSLE(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossSigmoidCrossEntropy(SDVariable labels, SDVariable logits, SDVariable weights, LossReduce lossReduce, double labelSmoothing) {
+        return new SigmoidCrossEntropyLoss(sameDiff(), lossReduce, logits, weights, labels, labelSmoothing).outputVariable();
     }
 
-
-    public SDVariable lossNegativeLogLikelihood(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
+    public SDVariable lossSoftmaxCrossEntropy(SDVariable labels, SDVariable logits, SDVariable weights, LossReduce lossReduce, double labelSmoothing) {
+        return new SoftmaxCrossEntropyLoss(sameDiff(), lossReduce, logits, weights, labels, labelSmoothing).outputVariable();
     }
 
 
-    public SDVariable lossPoisson(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
-
-    public SDVariable lossSquaredHinge(SDVariable iX, SDVariable i_y, int... dimensions) {
-        throw new UnsupportedOperationException();
-
-    }
-
     public SDVariable xwPlusB(SDVariable input, SDVariable weights, SDVariable bias) {
         return new XwPlusB(sameDiff(), input, weights, bias).outputVariable();
     }