Jiga228
109 lines
2.7 KiB
C++
109 lines
2.7 KiB
C++
#include <iostream>
|
|
#include <array>
|
|
#include <string>
|
|
|
|
#include "NeuronsNetworkBase.h"
|
|
|
|
using data_item = std::array<double, 7>;
|
|
|
|
class TitanicAI : public NeuronsNetworkBase<data_item>
|
|
{
|
|
constexpr static size_t count_neurons = 6;
|
|
protected:
|
|
double activation(double x) override
|
|
{
|
|
return 1/(1+exp(-x));
|
|
}
|
|
|
|
public:
|
|
TitanicAI() : NeuronsNetworkBase(count_neurons, 0.01)
|
|
{}
|
|
|
|
double call(const data_item& data) override
|
|
{
|
|
double sum = 0;
|
|
for (size_t i = 0; i < count_neurons; ++i)
|
|
{
|
|
sum += (*this)[i] * data[i + 1];
|
|
}
|
|
|
|
return activation(sum);
|
|
}
|
|
|
|
void train(double error, const data_item& data) override
|
|
{
|
|
for (size_t i = 0; i < count_neurons; ++i)
|
|
{
|
|
(*this)[i] += get_rate_train() * error * data[i + 1];
|
|
}
|
|
}
|
|
|
|
void print_masses()
|
|
{
|
|
for (size_t i = 0; i < count_neurons; ++i)
|
|
{
|
|
std::cout << (*this)[i] << " ";
|
|
}
|
|
std::cout << "\n";
|
|
}
|
|
};
|
|
|
|
constexpr size_t count_train_items = 700;
|
|
std::vector<data_item> train_data;
|
|
|
|
int main(int argc, char* argv[])
|
|
{
|
|
TitanicAI ai;
|
|
std::ifstream file("train.txt");
|
|
|
|
while (file.eof() == false)
|
|
{
|
|
data_item item{};
|
|
|
|
std::string vec;
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[0] = std::stod(vec);
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[1] = std::stod(vec);
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[2] = std::stod(vec);
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[3] = std::stod(vec);
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[4] = std::stod(vec);
|
|
file >> vec;
|
|
if (vec.empty() == false)
|
|
item[5] = std::stod(vec);
|
|
if (vec.empty() == false)
|
|
item[6] = std::stod(vec);
|
|
|
|
train_data.push_back(item);
|
|
}
|
|
|
|
for (size_t i = 0; i < count_train_items; ++i)
|
|
{
|
|
ai.train(train_data[i][0] - ai.call(train_data[i]), train_data[i]);
|
|
}
|
|
|
|
size_t score = 0;
|
|
const size_t max_score = train_data.size() - count_train_items;
|
|
constexpr double delta = 0.001;
|
|
for (size_t i = count_train_items + 1; i < train_data.size(); ++i)
|
|
{
|
|
double activate_result = ai.call(train_data[i]);
|
|
bool is_dead = activate_result + delta > 0.5;
|
|
if (is_dead == static_cast<bool>(train_data[i][0]))
|
|
++score;
|
|
}
|
|
|
|
std::cout << "Accuracy: " << score << '/' << max_score << " (float: " << (static_cast<double>(score) / static_cast<double>(max_score)) << ")\n";
|
|
std::cout << "Masses: "; ai.print_masses();
|
|
|
|
return 0;
|
|
}
|