new function to support LSTM Unit

include/NeuralNetwork/ActivationFunction/Linear.h

@@ -0,0 +1,26 @@
+#pragma once
+
+#include "./ActivationFunction.h"
+
+namespace NeuralNetwork {
+namespace ActivationFunction {
+
+	class Linear: public ActivationFunction {
+		public:
+			Linear(const float &lambdaP=1.0): lambda(lambdaP) {}
+			inline virtual float derivatedOutput(const float &,const float &) override { return lambda;  }
+			inline virtual float operator()(const float &x) override { return x*lambda; };
+
+			virtual ActivationFunction* clone() const override {
+				return new Linear(lambda);
+			}
+
+			virtual std::string stringify() const override {
+				return "{ \"class\": \"NeuralNetwork::ActivationFunction::Linear\", \"lamba\" : "+std::to_string(lambda)+"}";
+			}
+
+		protected:
+			float lambda;
+	};
+}
+}

include/NeuralNetwork/BasisFunction/BasisFunction.h

@@ -10,7 +10,7 @@ namespace BasisFunction {
 	class BasisFunction {
 		public:
 			virtual ~BasisFunction() {}
-			virtual float operator()(const std::vector<float>& weights, const std::vector<float>& input)=0;
+			virtual float operator()(const std::vector<float>& weights, const std::vector<float>& input) const =0;
 
 			/**
 			 * @brief Function returns clone of object

include/NeuralNetwork/BasisFunction/Linear.h

@@ -17,7 +17,7 @@ namespace BasisFunction {
 		public:
 			Linear() {}
 
-			inline virtual float computeStreaming(const std::vector<float>& weights, const std::vector<float>& input) override {
+			inline virtual float computeStreaming(const std::vector<float>& weights, const std::vector<float>& input) const override {
 				size_t inputSize=input.size();
 				size_t alignedPrev=inputSize-inputSize%4;
 
@@ -46,7 +46,7 @@ namespace BasisFunction {
 				return partialSolution.f[0];
 			}
 
-			inline virtual float compute(const std::vector<float>& weights, const std::vector<float>& input) override {
+			inline virtual float compute(const std::vector<float>& weights, const std::vector<float>& input) const override {
 				register float tmp = 0;
 				size_t inputSize=input.size();
 				for(size_t k=0;k<inputSize;k++) {

include/NeuralNetwork/BasisFunction/Product.h

@@ -0,0 +1,34 @@
+#pragma once
+
+#include "./BasisFunction.h"
+
+namespace NeuralNetwork {
+namespace BasisFunction {
+
+	class Product: public BasisFunction {
+		public:
+			Product() {}
+
+			/**
+			 * @brief function computes product of inputs, where weight > 0.5
+			 */
+			inline virtual float operator()(const std::vector<float>& weights, const std::vector<float>& input) const override {
+				float product=1.0;
+				for(size_t i=0;i<weights.size();i++) {
+					if(weights[i] > 0.5)
+						product=product*input[i];
+				}
+				return product;
+			}
+
+			virtual Product* clone() const override {
+				return new Product();
+			}
+
+			virtual std::string stringify() const override {
+				return "{ \"class\": \"NeuralNetwork::BasisFunction::Product\" }";
+			}
+
+	};
+}
+}

include/NeuralNetwork/BasisFunction/StreamingBasisFunction.h

@@ -13,11 +13,13 @@ namespace BasisFunction {
 				float f[4];
 			};
 
-			virtual float operator()(const std::vector<float>& weights, const std::vector<float>& input) override {
+			virtual float operator()(const std::vector<float>& weights, const std::vector<float>& input) const override {
 				return computeStreaming(weights,input);
 			}
-			virtual float computeStreaming(const std::vector<float>& weights, const std::vector<float>& input) =0;
+
-			virtual float compute(const std::vector<float>& weights, const std::vector<float>& input) =0;
+			virtual float computeStreaming(const std::vector<float>& weights, const std::vector<float>& input) const =0;
+
+			virtual float compute(const std::vector<float>& weights, const std::vector<float>& input) const =0;
 	};
 }
 }

tests/activation.cpp

@@ -1,6 +1,7 @@
 #include <NeuralNetwork/ActivationFunction/Heaviside.h>
 #include <NeuralNetwork/ActivationFunction/Sigmoid.h>
 #include <NeuralNetwork/ActivationFunction/HyperbolicTangent.h>
+#include <NeuralNetwork/ActivationFunction/Linear.h>
 
 #include <iostream>
 #include <cassert>
@@ -26,14 +27,14 @@ int main() {
 
 	{
 		NeuralNetwork::ActivationFunction::Sigmoid s(0.7);
-		assert(s(0.1) > 0.517483);
+		assert(s(0.1) > 0.482407);
-		assert(s(0.1) < 0.51750);
+		assert(s(0.1) < 0.482607);
 
-		assert(s(10) > 0.998989);
+		assert(s(10) > 0.000901051);
-		assert(s(10) < 0.999189);
+		assert(s(10) < 0.000921051);
 	}
 	{
-		NeuralNetwork::ActivationFunction::Sigmoid s(5);
+		NeuralNetwork::ActivationFunction::Sigmoid s(-5);
 		assert(s(0.1) > 0.622359);
 		assert(s(0.1) < 0.622559);
 		
@@ -41,7 +42,7 @@ int main() {
 		assert(s(0.7) < 0.970788);
 	}
 	{
-		NeuralNetwork::ActivationFunction::Sigmoid s(0.7);
+		NeuralNetwork::ActivationFunction::Sigmoid s(-0.7);
 		U.a[0]=0.1;
 		U.a[1]=10;
 		U.v=s(U.v);
@@ -52,8 +53,20 @@ int main() {
 		assert(U.a[1] > 0.998989);
 		assert(U.a[1] < 0.999189);
 	}
+	{
+		NeuralNetwork::ActivationFunction::Linear s(1.0);
+		assert(s(0.5) > 0.4999);
+		assert(s(0.5) < 0.5001);
 
-	std::cout << "OK" << std::endl;
+		assert(s(0.0) == 0.0);
+	}
+	{
+		NeuralNetwork::ActivationFunction::Linear s(0.7);
+		assert(s(0.0) == 0.0);
+
+		assert(s(1.0) > 0.6999);
+		assert(s(1.0) < 0.7001);
+	}
 
 	return 0;
 }

tests/basis.cpp

@@ -1,4 +1,5 @@
 #include <NeuralNetwork/BasisFunction/Linear.h>
+#include <NeuralNetwork/BasisFunction/Product.h>
 
 #include <iostream>
 #include <cassert>
@@ -33,6 +34,14 @@ int main() {
 		assert(220.0==l.computeStreaming(w,w));
 		assert(220.0==l.compute(w,w));
 	}
+	{
+		NeuralNetwork::BasisFunction::Product l;
+		std::vector<float> w({0,0.501,1});
+		std::vector<float> i({0,0.2,0.3});
+
+		assert(l(w,i) > 0.05999);
+		assert(l(w,i) < 0.06001);
+	}
 /*
 	std::vector<float> w;
 	std::vector<float> i;
@@ -61,6 +70,5 @@ int main() {
 		std::cout << "SSE :" << diff.count() << " s\n";
 	}
 */
-	std::cout <<"OK" << std::endl;
 
 }