/// <summary> /// Connect layers from a BasicNetwork. Used internally only. /// </summary> /// <param name="network">The BasicNetwork.</param> /// <param name="fromLayerIdx">The from layer index.</param> /// <param name="source">The from layer.</param> /// <param name="target">The target.</param> private void ConnectLayersFromBasic(BasicNetwork network, int fromLayerIdx, IFreeformLayer source, IFreeformLayer target) { for (int targetNeuronIdx = 0; targetNeuronIdx < target.Count; targetNeuronIdx++) { for (int sourceNeuronIdx = 0; sourceNeuronIdx < source.Count; sourceNeuronIdx++) { IFreeformNeuron sourceNeuron = source.Neurons[sourceNeuronIdx]; IFreeformNeuron targetNeuron = target.Neurons[targetNeuronIdx]; // neurons with no input (i.e. bias neurons) if (targetNeuron.InputSummation == null) { continue; } IFreeformConnection connection = _connectionFactory .Factor(sourceNeuron, targetNeuron); sourceNeuron.AddOutput(connection); targetNeuron.AddInput(connection); double weight = network.GetWeight(fromLayerIdx, sourceNeuronIdx, targetNeuronIdx); connection.Weight = weight; } } }
/// <inheritdoc/> protected override void LearnConnection(IFreeformConnection connection) { double gradient = connection.GetTempTraining(0); double delta = (gradient * _learningRate) + (connection.GetTempTraining(1) * _momentum); connection.SetTempTraining(1, delta); connection.Weight += delta; }
/// <summary> /// Connect two layers. /// </summary> /// <param name="source">The source layer.</param> /// <param name="target">The target layer.</param> /// <param name="theActivationFunction">The activation function to use.</param> /// <param name="biasActivation">The bias activation to use.</param> /// <param name="isRecurrent">True, if this is a recurrent connection.</param> public void ConnectLayers(IFreeformLayer source, IFreeformLayer target, IActivationFunction theActivationFunction, double biasActivation, bool isRecurrent) { // create bias, if requested if (biasActivation > EncogFramework.DefaultDoubleEqual) { // does the source already have a bias? if (source.HasBias) { throw new FreeformNetworkError( "The source layer already has a bias neuron, you cannot create a second."); } IFreeformNeuron biasNeuron = _neuronFactory .FactorRegular(null); biasNeuron.Activation = biasActivation; biasNeuron.IsBias = true; source.Add(biasNeuron); } // create connections foreach (IFreeformNeuron targetNeuron in target.Neurons) { // create the summation for the target IInputSummation summation = targetNeuron.InputSummation; // do not create a second input summation if (summation == null) { summation = _summationFactory.Factor(theActivationFunction); targetNeuron.InputSummation = summation; } // connect the source neurons to the target neuron foreach (IFreeformNeuron sourceNeuron in source.Neurons) { IFreeformConnection connection = _connectionFactory .Factor(sourceNeuron, targetNeuron); sourceNeuron.AddOutput(connection); targetNeuron.AddInput(connection); } } }
/// <inheritdoc/> public void Add(IFreeformConnection connection) { _inputs.Add(connection); }
/// <inheritdoc /> protected override void LearnConnection(IFreeformConnection connection) { // multiply the current and previous gradient, and take the // sign. We want to see if the gradient has changed its sign. int change = EncogMath .Sign(connection .GetTempTraining(TempGradient) *connection .GetTempTraining(TempLastGradient)); double weightChange = 0; // if the gradient has retained its sign, then we increase the // delta so that it will converge faster if (change > 0) { double delta = connection .GetTempTraining(TempUpdate) *RPROPConst.PositiveEta; delta = Math.Min(delta, _maxStep); weightChange = EncogMath .Sign(connection .GetTempTraining(TempGradient)) *delta; connection.SetTempTraining( TempUpdate, delta); connection .SetTempTraining( TempLastGradient, connection .GetTempTraining(TempGradient)); } else if (change < 0) { // if change<0, then the sign has changed, and the last // delta was too big double delta = connection .GetTempTraining(TempUpdate) *RPROPConst.NegativeEta; delta = Math.Max(delta, RPROPConst.DeltaMin); connection.SetTempTraining( TempUpdate, delta); weightChange = -connection .GetTempTraining(TempLastWeightDelta); // set the previous gradient to zero so that there will be no // adjustment the next iteration connection.SetTempTraining( TempLastGradient, 0); } else if (change == 0) { // if change==0 then there is no change to the delta double delta = connection .GetTempTraining(TempUpdate); weightChange = EncogMath .Sign(connection .GetTempTraining(TempGradient)) *delta; connection .SetTempTraining( TempLastGradient, connection .GetTempTraining(TempGradient)); } // apply the weight change, if any connection.Weight += weightChange; connection.SetTempTraining( TempLastWeightDelta, weightChange); }
/// <inheritdoc /> protected override void LearnConnection(IFreeformConnection connection) { // multiply the current and previous gradient, and take the // sign. We want to see if the gradient has changed its sign. int change = EncogMath .Sign(connection .GetTempTraining(TempGradient) * connection .GetTempTraining(TempLastGradient)); double weightChange = 0; // if the gradient has retained its sign, then we increase the // delta so that it will converge faster if (change > 0) { double delta = connection .GetTempTraining(TempUpdate) * RPROPConst.PositiveEta; delta = Math.Min(delta, _maxStep); weightChange = EncogMath .Sign(connection .GetTempTraining(TempGradient)) * delta; connection.SetTempTraining( TempUpdate, delta); connection .SetTempTraining( TempLastGradient, connection .GetTempTraining(TempGradient)); } else if (change < 0) { // if change<0, then the sign has changed, and the last // delta was too big double delta = connection .GetTempTraining(TempUpdate) * RPROPConst.NegativeEta; delta = Math.Max(delta, RPROPConst.DeltaMin); connection.SetTempTraining( TempUpdate, delta); weightChange = -connection .GetTempTraining(TempLastWeightDelta); // set the previous gradient to zero so that there will be no // adjustment the next iteration connection.SetTempTraining( TempLastGradient, 0); } else if (change == 0) { // if change==0 then there is no change to the delta double delta = connection .GetTempTraining(TempUpdate); weightChange = EncogMath .Sign(connection .GetTempTraining(TempGradient)) * delta; connection .SetTempTraining( TempLastGradient, connection .GetTempTraining(TempGradient)); } // apply the weight change, if any connection.Weight += weightChange; connection.SetTempTraining( TempLastWeightDelta, weightChange); }
/// <inheritdoc/> public void AddOutput(IFreeformConnection connection) { _outputConnections.Add(connection); }
/// <inheritdoc/> public void AddInput(IFreeformConnection connection) { InputSummation.Add(connection); }
/// <inheritdoc/> public void Add(IFreeformConnection connection) { _inputs.Add(connection); }
/// <inheritdoc/> public void AddOutput(IFreeformConnection connection) { _outputConnections.Add(connection); }
/// <inheritdoc/> public void AddInput(IFreeformConnection connection) { InputSummation.Add(connection); }
/// <summary> /// Learn for a single connection. /// </summary> /// <param name="connection">The connection to learn from.</param> protected abstract void LearnConnection(IFreeformConnection connection);
/// <inheritdoc/> protected override void LearnConnection(IFreeformConnection connection) { double gradient = connection.GetTempTraining(0); double delta = (gradient * _learningRate) + (connection.GetTempTraining(1) * _momentum); connection.SetTempTraining(1, delta); connection.Weight += delta; }
/// <summary> /// Learn for a single connection. /// </summary> /// <param name="connection">The connection to learn from.</param> protected abstract void LearnConnection(IFreeformConnection connection);