public void Main()
{
const int gameSize = 3;
var game = new Game(gameSize);
var q_table = Matematica.ArrayJagged_Ini <double>(gameSize * gameSize, 4);
int num_episodes = 10000;
int max_steps_per_episode = 100;
double learning_rate = 0.1;
double discount_rate = 0.99;
double exploration_rate = 1;
double max_exploration_rate = 1;
double min_exploration_rate = 0.01;
double exploration_decay_rate = 0.01;
var rewards_all_episodes = new List <int>();
//inside vars
for (int i = 0; i < num_episodes; i++)
{
var state = game.GameReset();
int rewards_current_episode = 0;
for (int e = 0; e < max_steps_per_episode; e++)
{
// Exploration-exploitation trade-off
double exploration_rate_threshold = Matematica.rng.NextDouble();
int action = Matematica.ArgMax(q_table[state]);
if (exploration_rate_threshold <= exploration_rate)
{
action = Explore_action(action);
}
//(new_state, reward, done, info) = env.step(action);
var(new_state, reward, done) = game.Step(action);
q_table[state][action] = Matematica.Q_Table(q_table[state][action], q_table[new_state], learning_rate, reward, discount_rate);
state = new_state;
rewards_current_episode += reward;
if (done)
{
break;
}
}
exploration_rate = Matematica.Exploration(min_exploration_rate, max_exploration_rate, exploration_decay_rate, i);
rewards_all_episodes.Add(rewards_current_episode);
}
Average_Reward_Print(rewards_all_episodes, num_episodes);
Console.WriteLine(q_table[2][1]);
}
public (int new_state, int reward, bool done) Step(int action)
{
MoveByIndex(action);
int new_state = Matematica.Array2D_GetFlatIndex(game_matriz, pos_y, pos_x);
var reward = rewards[pos_x, pos_y];
totalReward += reward;
var done = false;
if (rewards_to_win <= totalReward)
{
done = true;
}
return(new_state, reward, done);
}
public int GameReset()
{
ReSet(border);
return(Matematica.Array2D_GetFlatIndex(game_matriz, pos_y, pos_x));
}
