fix: resolve all build errors in nested learning implementation

Fixed multiple compilation errors across HopeNetwork and ContinuumMemorySystemLayer:

- Fixed RecurrentLayer and DenseLayer constructor calls to use correct signatures
- Added explicit casts to resolve activation function constructor ambiguity
- Fixed LayerBase constructor to use 2-parameter version
- Initialized non-nullable fields in HopeNetwork constructor
- Added null coalescing for nullable loss function parameter
- Replaced ILossFunction method names (ComputeLoss → CalculateLoss)
- Fixed Vector construction to use AiDotNet pattern instead of MathNet.Numerics
- Fixed Tensor construction using correct constructor signature
- Replaced protected Parameters access with public ParameterCount and GetParameters()
- Added LastInput and LastOutput fields to ContinuumMemorySystemLayer
- Fixed test ambiguity in Forward() calls with explicit type casts

Build now completes successfully with 0 errors.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: ooples

src/NestedLearning/HopeNetwork.cs

@@ -39,7 +39,7 @@ public HopeNetwork(
         int numCMSLevels = 4,
         int numRecurrentLayers = 3,
         int inContextLearningLevels = 5)
-        : base(architecture, lossFunction, maxGradNorm: 1.0)
+        : base(architecture, lossFunction ?? new MeanSquaredErrorLoss<T>(), maxGradNorm: 1.0)
     {
         _hiddenDim = hiddenDim;
         _numCMSLevels = numCMSLevels;
@@ -48,6 +48,10 @@ public HopeNetwork(
         _adaptationStep = 0;
         _selfModificationRate = _numOps.FromDouble(0.01);
 
+        // Initialize arrays to avoid non-nullable warnings
+        _cmsBlocks = new ContinuumMemorySystemLayer<T>[numCMSLevels];
+        _recurrentLayers = new RecurrentLayer<T>[numRecurrentLayers];
+
         // Initialize context flow for multi-level optimization
         _contextFlow = new ContextFlow<T>(hiddenDim, inContextLearningLevels);
 
@@ -79,9 +83,9 @@ protected override void InitializeLayers()
         for (int i = 0; i < _numRecurrentLayers; i++)
         {
             _recurrentLayers[i] = new RecurrentLayer<T>(
-                inputShape: new[] { _hiddenDim },
-                outputUnits: _hiddenDim,
-                activation: ActivationFunction.Tanh);
+                _hiddenDim,
+                _hiddenDim,
+                (IActivationFunction<T>)new TanhActivation<T>());
 
             Layers.Add(_recurrentLayers[i]);
         }
@@ -330,11 +334,16 @@ public void ResetRecurrentState()
     /// </summary>
     public void AddOutputLayer(int outputDim, ActivationFunction activation = ActivationFunction.Linear)
     {
-        _outputLayer = new DenseLayer<T>(
-            inputShape: new[] { _hiddenDim },
-            outputUnits: outputDim,
-            activation: activation);
+        IActivationFunction<T> activationFunc = activation switch
+        {
+            ActivationFunction.Tanh => new TanhActivation<T>(),
+            ActivationFunction.Softmax => new SoftmaxActivation<T>(),
+            ActivationFunction.Sigmoid => new SigmoidActivation<T>(),
+            ActivationFunction.ReLU => new ReLUActivation<T>(),
+            _ => new IdentityActivation<T>()
+        };
 
+        _outputLayer = new DenseLayer<T>(_hiddenDim, outputDim, activationFunc);
         Layers.Add(_outputLayer);
     }
 
@@ -451,11 +460,18 @@ public override void Train(Tensor<T> input, Tensor<T> expectedOutput)
         // Forward pass
         var prediction = Forward(input);
 
+        // Convert tensors to vectors for loss computation
+        var predictionVector = new Vector<T>(prediction.ToArray());
+        var expectedVector = new Vector<T>(expectedOutput.ToArray());
+
         // Compute loss
-        var loss = LossFunction.ComputeLoss(prediction, expectedOutput);
+        var loss = LossFunction.CalculateLoss(predictionVector, expectedVector);
 
         // Compute loss gradient
-        var lossGradient = LossFunction.ComputeGradient(prediction, expectedOutput);
+        var lossGradientVector = LossFunction.CalculateDerivative(predictionVector, expectedVector);
+
+        // Convert gradient vector back to tensor for backward pass
+        var lossGradient = new Tensor<T>(prediction.Shape, lossGradientVector);
 
         // Backward pass
         Backward(lossGradient);
@@ -503,7 +519,7 @@ public override ModelMetadata<T> GetModelMetadata()
             { "RecurrentLayers", _numRecurrentLayers },
             { "InContextLearningLevels", _inContextLearningLevels },
             { "AdaptationStep", _adaptationStep },
-            { "SelfModificationRate", _selfModificationRate },
+            { "SelfModificationRate", (object?)_selfModificationRate ?? 0 },
             { "ParameterCount", ParameterCount },
             { "LayerCount", Layers?.Count ?? 0 }
         };

src/NeuralNetworks/Layers/ContinuumMemorySystemLayer.cs

@@ -22,6 +22,8 @@ public class ContinuumMemorySystemLayer<T> : LayerBase<T>
     private readonly int[] _stepCounters;
     private readonly Vector<T>[] _storedInputs;  // Store input to each MLP block for Modified GD
     private int _globalStep;
+    private Tensor<T>? LastInput;
+    private Tensor<T>? LastOutput;
     private static readonly INumericOperations<T> _numOps = MathHelper.GetNumericOperations<T>();
 
     /// <summary>
@@ -43,7 +45,7 @@ public ContinuumMemorySystemLayer(
         int numFrequencyLevels = 3,
         int[]? updateFrequencies = null,
         T[]? learningRates = null)
-        : base(inputShape, new[] { hiddenDim }, null, null)
+        : base(inputShape, new[] { hiddenDim })
     {
         // Validate inputs
         if (inputShape == null || inputShape.Length == 0)
@@ -103,10 +105,7 @@ public ContinuumMemorySystemLayer(
 
         for (int i = 0; i < numFrequencyLevels; i++)
         {
-            _mlpBlocks[i] = new DenseLayer<T>(
-                inputShape: new[] { currentDim },
-                outputUnits: hiddenDim,
-                activation: ActivationFunction.ReLU);
+            _mlpBlocks[i] = new DenseLayer<T>(currentDim, hiddenDim, (IActivationFunction<T>)new ReLUActivation<T>());
             currentDim = hiddenDim;
         }
 
@@ -115,7 +114,7 @@ public ContinuumMemorySystemLayer(
         _stepCounters = new int[numFrequencyLevels];
         for (int i = 0; i < numFrequencyLevels; i++)
         {
-            int paramCount = _mlpBlocks[i].Parameters.Length;
+            int paramCount = _mlpBlocks[i].ParameterCount;
             _accumulatedGradients[i] = new Vector<T>(paramCount);
             _stepCounters[i] = 0;
         }
@@ -238,7 +237,7 @@ private void UpdateLevelParameters(int level)
         if (_mlpBlocks[level] == null)
             throw new InvalidOperationException($"MLP block at level {level} is null");
 
-        var currentParams = _mlpBlocks[level].Parameters;
+        var currentParams = _mlpBlocks[level].GetParameters();
         if (currentParams == null || currentParams.Length == 0)
             throw new InvalidOperationException($"MLP block at level {level} has no parameters");
 
@@ -295,8 +294,8 @@ public void ConsolidateMemory()
             if (_mlpBlocks[i + 1] == null)
                 throw new InvalidOperationException($"MLP block at level {i + 1} is null");
 
-            var fastParams = _mlpBlocks[i].Parameters;
-            var slowParams = _mlpBlocks[i + 1].Parameters;
+            var fastParams = _mlpBlocks[i].GetParameters();
+            var slowParams = _mlpBlocks[i + 1].GetParameters();
 
             if (fastParams == null || fastParams.Length == 0)
                 throw new InvalidOperationException($"Fast MLP at level {i} has no parameters");
@@ -404,7 +403,7 @@ public override Vector<T> GetParameters()
             if (mlp == null)
                 throw new InvalidOperationException("MLP block is null");
 
-            totalParams += mlp.Parameters.Length;
+            totalParams += mlp.ParameterCount;
         }
 
         // Concatenate all parameters
@@ -413,7 +412,7 @@ public override Vector<T> GetParameters()
 
         foreach (var mlp in _mlpBlocks)
         {
-            var mlpParams = mlp.Parameters;
+            var mlpParams = mlp.GetParameters();
             for (int i = 0; i < mlpParams.Length; i++)
             {
                 allParams[offset + i] = mlpParams[i];
@@ -444,7 +443,7 @@ public override void SetParameters(Vector<T> parameters)
             if (mlp == null)
                 throw new InvalidOperationException("MLP block is null");
 
-            totalParams += mlp.Parameters.Length;
+            totalParams += mlp.ParameterCount;
         }
 
         if (parameters.Length != totalParams)
@@ -458,7 +457,7 @@ public override void SetParameters(Vector<T> parameters)
         int offset = 0;
         foreach (var mlp in _mlpBlocks)
         {
-            int mlpParamCount = mlp.Parameters.Length;
+            int mlpParamCount = mlp.ParameterCount;
             var mlpParams = new Vector<T>(mlpParamCount);
 
             for (int i = 0; i < mlpParamCount; i++)
@@ -499,7 +498,7 @@ public override Vector<T> GetParameterGradients()
             if (mlp == null)
                 throw new InvalidOperationException("MLP block is null");
 
-            totalParams += mlp.Parameters.Length;
+            totalParams += mlp.ParameterCount;
         }
 
         // Concatenate all accumulated gradients

tests/AiDotNet.Tests/UnitTests/NestedLearning/ContinuumMemorySystemLayerTests.cs

@@ -173,7 +173,7 @@ public void Forward_WithNullInput_ThrowsArgumentNullException()
                 numFrequencyLevels: 3);
 
             // Act & Assert
-            Assert.Throws<ArgumentNullException>(() => layer.Forward(null!));
+            Assert.Throws<ArgumentNullException>(() => layer.Forward((Tensor<double>)null!));
         }
 
         [Fact]
@@ -202,7 +202,7 @@ public void Forward_ProcessesSequentiallyThroughAllMLPBlocks()
             foreach (var block in mlpBlocks)
             {
                 Assert.NotNull(block);
-                Assert.True(block.Parameters.Length > 0);
+                Assert.True(block.ParameterCount > 0);
             }
         }
 
@@ -302,7 +302,7 @@ public void ConsolidateMemory_TransfersKnowledgeBetweenLevels()
             foreach (var block in mlpBlocks)
             {
                 Assert.NotNull(block);
-                Assert.True(block.Parameters.Length > 0);
+                Assert.True(block.ParameterCount > 0);
             }
         }