diff --git a/src/NeuralNetwork/Learning/QuickPropagation.cpp b/src/NeuralNetwork/Learning/QuickPropagation.cpp
index b943d48..bf9f789 100644
--- a/src/NeuralNetwork/Learning/QuickPropagation.cpp
+++ b/src/NeuralNetwork/Learning/QuickPropagation.cpp
@@ -5,6 +5,8 @@
 
 void NeuralNetwork::Learning::QuickPropagation::updateWeights(const std::vector<float> &input) {
 
+	float shrinkFactor=_maxChange/(_maxChange+1.0);
+
 	for(std::size_t layerIndex=1;layerIndex<network.size();layerIndex++) {
 		auto &layer=network[layerIndex];
 		auto &prevLayer=network[layerIndex-1];
@@ -14,13 +16,22 @@ void NeuralNetwork::Learning::QuickPropagation::updateWeights(const std::vector<
 
 		for(std::size_t j=1;j<layerSize;j++)  {
 
-			float newChange=slopes[layerIndex][j]/(previousSlopes[layerIndex][j]-slopes[layerIndex][j]) * lastWeightChange[layerIndex][j];
+			float newChange=0;
 
-			// according to original paper
-			newChange+= slopes[layerIndex][j]*_epsilon;
+			if(fabs (lastWeightChange[layerIndex][j])> 0.0001) {
+				if(std::signbit(lastWeightChange[layerIndex][j]) == std::signbit(slopes[layerIndex][j])) {
+					newChange+= slopes[layerIndex][j]*_epsilon;
 
-			if(newChange > lastWeightChange[layerIndex][j]*_maxChange) {
-				newChange=lastWeightChange[layerIndex][j];
+					if(fabs(slopes[layerIndex][j]) > fabs(shrinkFactor * previousSlopes[layerIndex][j])) {
+						newChange += _maxChange * lastWeightChange[layerIndex][j];
+					}else {
+						newChange+=slopes[layerIndex][j]/(previousSlopes[layerIndex][j]-slopes[layerIndex][j]) * lastWeightChange[layerIndex][j];
+					}
+				} else {
+					newChange+=slopes[layerIndex][j]/(previousSlopes[layerIndex][j]-slopes[layerIndex][j]) * lastWeightChange[layerIndex][j];
+				}
+			} else {
+				newChange+= slopes[layerIndex][j]*_epsilon;
 			}
 
 			weightChange[layerIndex][j]=newChange;