shin/NeuralNetworkLib
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Modified FeedForward to allow set activation to whole Layer and added XOR test for FF

Browse Source
This commit is contained in:
Tomáš Černík
2016-02-03 21:16:35 +01:00
parent ea4ce22867
commit 567fcd2373
5 changed files with 64 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/NeuralNetwork/FeedForward/Layer.h
View File

@@ -16,10 +16,10 @@ namespace FeedForward {
class Layer : public Stringifiable {
public:
Layer(std::size_t size = 0):neurons() {
Layer(std::size_t size, const ActivationFunction::ActivationFunction &activationFunction):neurons() {
neurons.push_back(new BiasNeuron);
for(std::size_t i=0;i<size;i++) {
neurons.push_back(new Neuron(neurons.size()));
neurons.push_back(new Neuron(neurons.size(),activationFunction));
}
}

37
include/NeuralNetwork/FeedForward/Network.h
View File

@@ -9,6 +9,8 @@
#include <iomanip>
#include <limits>
#include <iostream>
namespace NeuralNetwork {
namespace FeedForward {
@@ -29,24 +31,27 @@ namespace FeedForward {
appendLayer(_inputSize);
};
Layer& appendLayer(std::size_t size=1) {
layers.push_back(Layer(size));
if(layers.size() > 1)
layers.back().setInputSize(layers[layers.size()-2].size());
return layers.back();
}
Layer& operator[](const std::size_t &id) {
return layers[id];
}
/**
* @brief Virtual destructor for Network
*/
virtual ~Network() {
};
for(auto &layer:layers) {
delete layer;
}
}
Layer& appendLayer(std::size_t size=1, const ActivationFunction::ActivationFunction &activationFunction=ActivationFunction::Sigmoid(-4.9)) {
layers.push_back(new Layer(size,activationFunction));
if(layers.size() > 1)
layers.back()->setInputSize(layers[layers.size()-2]->size());
return *layers[layers.size()-1];//.back();
}
Layer& operator[](const std::size_t &id) {
return *layers[id];
}
/**
* @brief This is a function to compute one iterations of network
@@ -66,7 +71,7 @@ namespace FeedForward {
if(!first) {
out << ",";
}
out << layer;
out << *layer;
first=false;
}
out << "]";
@@ -74,7 +79,7 @@ namespace FeedForward {
}
protected:
std::vector<Layer> layers;
std::vector<Layer*> layers;
};
}
}

7
include/NeuralNetwork/Neuron.h
View File

@@ -92,9 +92,10 @@ namespace NeuralNetwork
class Neuron: public NeuronInterface
{
public:
Neuron(unsigned long _id=0): NeuronInterface(), basis(new BasisFunction::Linear),
activation(new ActivationFunction::Sigmoid(-4.9)),
id_(_id),weights(_id+1),_output(0),_value(0) {
Neuron(unsigned long _id=0, const ActivationFunction::ActivationFunction &activationFunction=ActivationFunction::Sigmoid(-4.9)):
NeuronInterface(), basis(new BasisFunction::Linear),
activation(activationFunction.clone()),
id_(_id),weights(_id+1),_output(0),_value(0) {
}
Neuron(const Neuron &r): NeuronInterface(), basis(r.basis->clone()), activation(r.activation->clone()),id_(r.id_),

6
src/NeuralNetwork/FeedForward/Network.cpp
View File

@@ -9,16 +9,16 @@ std::vector<float> NeuralNetwork::FeedForward::Network::computeOutput(const std:
// 0 is bias
partial1[0]=1.0;
for(int i=0;i<input.size();i++) {
for(std::size_t i=0;i<input.size();i++) {
partial1[i+1]=input[i];
}
for(std::size_t i=1;i<layers.size();i++) {
layers[i].solve(*partialInputPtr,*partialOutputPtr);
layers[i]->solve(*partialInputPtr,*partialOutputPtr);
std::swap(partialInputPtr,partialOutputPtr);
}
for(int i=0;i<partialInputPtr->size()-1;i++) {
for(std::size_t i=0;i<partialInputPtr->size()-1;i++) {
partialInputPtr->operator[](i)=partialInputPtr->operator[](i+1);
}

45
tests/feedforward.cpp
View File

@@ -1,5 +1,6 @@
#include <NeuralNetwork/FeedForward/Network.h>
#include <cassert>
#include <iostream>
void printVec(const std::vector<float> &v) {
@@ -10,22 +11,44 @@ void printVec(const std::vector<float> &v) {
}
int main() {
NeuralNetwork::FeedForward::Network n(2);
{ // XOR problem
NeuralNetwork::FeedForward::Network n(2);
NeuralNetwork::ActivationFunction::Sigmoid a(-1);
NeuralNetwork::FeedForward::Layer &hidden=n.appendLayer(2,a);
NeuralNetwork::FeedForward::Layer &out = n.appendLayer(1,a);
NeuralNetwork::FeedForward::Layer &sec=n.appendLayer(4);
hidden[1].setWeight(n[0][0],7);
hidden[1].setWeight(n[0][1],-4.7);
hidden[1].setWeight(n[0][2],-4.7);
NeuralNetwork::FeedForward::Layer &in = n[0];
hidden[2].setWeight(n[0][0],2.6);
hidden[2].setWeight(n[0][1],-6.4);
hidden[2].setWeight(n[0][2],-6.4);
NeuralNetwork::FeedForward::Layer &t = n.appendLayer(4);
sec[1].setWeight(in[1],-1.0);
out[1].setWeight(hidden[0],-4.5);
out[1].setWeight(hidden[1],9.6);
out[1].setWeight(hidden[2],-6.8);
sec[1].setWeight(in[2],-1.0);
sec[2].setWeight(in[2],-1.0);
t[2].setWeight(sec[2],-1.0);
{
std::vector<float> ret =n.computeOutput({1,1});
assert(ret[0] < 0.5);
}
std::vector<float> ret =n.computeOutput({0.7,0.7});
printVec(ret);
std::cout << n;
{
std::vector<float> ret =n.computeOutput({0,1});
assert(ret[0] > 0.5);
}
{
std::vector<float> ret =n.computeOutput({1,0});
assert(ret[0] > 0.5);
}
{
std::vector<float> ret =n.computeOutput({0,0});
assert(ret[0] < 0.5);
}
}
}