Merge pull request #137 from cbovar/Remove_Unused_Regularization

Remove unused regularization parameters.
cbovar · Feb 10, 2019 · 74716a6 · 74716a6
2 parents 7f58b36 + 8a1252d
commit 74716a6
Show file tree

Hide file tree

Showing 7 changed files with 2 additions and 20 deletions.
diff --git a/Examples/Classify2DDemo/Program.cs b/Examples/Classify2DDemo/Program.cs
@@ -52,7 +52,7 @@ private static void Classify2DDemo()
  labels.Add(1);
  var n = labels.Count;
 
- var trainer = new SgdTrainer<double>(net) { LearningRate = 0.01, L2Decay = 0.001, BatchSize = n };
+ var trainer = new SgdTrainer<double>(net) { LearningRate = 0.01, BatchSize = n };
 
  // Training
  do

diff --git a/Examples/FluentMnistDemo/Program.cs b/Examples/FluentMnistDemo/Program.cs
@@ -45,7 +45,6 @@ private void MnistDemo()
  {
  LearningRate = 0.01,
  BatchSize = 20,
- L2Decay = 0.001,
  Momentum = 0.9
  };
 

diff --git a/Examples/MinimalExample/Program.cs b/Examples/MinimalExample/Program.cs
@@ -39,7 +39,7 @@ private static void Main()
  // prob is a Volume. Volumes have a property Weights that stores the raw data, and WeightGradients that stores gradients
  Console.WriteLine("probability that x is class 0: " + prob.Get(0)); // prints e.g. 0.50101
 
- var trainer = new SgdTrainer(net) { LearningRate = 0.01, L2Decay = 0.001 };
+ var trainer = new SgdTrainer(net) { LearningRate = 0.01 };
  trainer.Train(x, BuilderInstance.Volume.From(new[] { 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 }, new Shape(1, 1, 10, 1))); // train the network, specifying that x is class zero
 
  var prob2 = net.Forward(x);

diff --git a/Examples/MnistDemo/Program.cs b/Examples/MnistDemo/Program.cs
@@ -45,7 +45,6 @@ private void MnistDemo()
  {
  LearningRate = 0.01,
  BatchSize = 20,
- L2Decay = 0.001,
  Momentum = 0.9
  };
 

diff --git a/src/ConvNetSharp.Core/Layers/ConvLayer.cs b/src/ConvNetSharp.Core/Layers/ConvLayer.cs
@@ -13,9 +13,6 @@ public class ConvLayer<T> : LayerBase<T>, IDotProductLayer<T> where T : struct,
 
  public ConvLayer(Dictionary<string, object> data) : base(data)
  {
- this.L1DecayMul = Ops<T>.Zero;
- this.L2DecayMul = Ops<T>.One;
-
  this.FilterCount = Convert.ToInt32(data["FilterCount"]);
  this.Width = Convert.ToInt32(data["Width"]);
  this.Height = Convert.ToInt32(data["Height"]);
@@ -32,9 +29,6 @@ public ConvLayer(Dictionary<string, object> data) : base(data)
 
  public ConvLayer(int width, int height, int filterCount)
  {
- this.L1DecayMul = Ops<T>.Zero;
- this.L2DecayMul = Ops<T>.One;
-
  this.FilterCount = filterCount;
  this.Width = width;
  this.Height = height;
@@ -54,10 +48,6 @@ public ConvLayer(int width, int height, int filterCount)
 
  public int FilterCount { get; }
 
- public T L1DecayMul { get; set; }
-
- public T L2DecayMul { get; set; }
-
  public int Stride
  {
  get { return this._stride; }

diff --git a/src/ConvNetSharp.Core/Training/SgdTrainer.cs b/src/ConvNetSharp.Core/Training/SgdTrainer.cs
@@ -18,10 +18,6 @@ public SgdTrainer(INet<T> net) : base(net)
  {
  }
 
- public T L1Decay { get; set; }
-
- public T L2Decay { get; set; }
-
  public T Momentum { get; set; }
 
  public T LearningRate { get; set; }

diff --git a/src/ConvNetSharp.Performance.Tests/Program.cs b/src/ConvNetSharp.Performance.Tests/Program.cs
@@ -117,8 +117,6 @@ private static void ExecuteNeuralNet(
  var trainer = new SgdTrainer(net);
  trainer.LearningRate = 0.01;
  trainer.Momentum = 0.5;
- trainer.L1Decay = 0.01;
- trainer.L2Decay = 0.01;
  trainer.BatchSize = batchSize;
 
  for (var i = 0; i < iterations; i++)