本文整理汇总了C++中Environment::act方法的典型用法代码示例。如果您正苦于以下问题：C++ Environment::act方法的具体用法？C++ Environment::act怎么用？C++ Environment::act使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类Environment的用法示例。
在下文中一共展示了Environment::act方法的2个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的C++代码示例。
示例1: evaluatePolicy
double HumanAgent::evaluatePolicy(Environment<bool>& env){
#ifdef __USE_SDL
Action action;
int reward = 0;
int totalReward = 0;
int cumulativeReward = 0;
//Repeat (for each episode):
for(int episode = 0; episode < numEpisodesToEval; episode++){
int step = 0;
while(!env.game_over() && step < maxStepsInEpisode) {
action = receiveAction();
//If one wants to save trajectories, this is where the trajectory is saved:
if(toSaveTrajectory){
saveTrajectory(action);
}
reward = env.act(action);
cumulativeReward += reward;
step++;
}
printf("Episode %d, Cumulative Reward: %d\n", episode + 1, cumulativeReward);
totalReward += cumulativeReward;
cumulativeReward = 0;
env.reset_game(); //Start the game again when the episode is over
}
return double(totalReward)/numEpisodesToEval;
}示例2: evaluatePolicy
double RandomAgent::evaluatePolicy(Environment<bool>& env){
int reward = 0;
int totalReward = 0;
int cumulativeReward = 0;
int numActions;
ActionVect actions;
//Check if one wants to sample from all possible actions or only the valid ones:
if(useMinActions){
actions = env.getMinimalActionSet();
}
else{
actions = env.getLegalActionSet();
}
numActions = actions.size();
printf("Number of Actions: %d\n\n", numActions);
//Repeat (for each episode):
for(int episode = 0; episode < numEpisodesToEval; episode++){
int step = 0;
while(!env.game_over() && step < maxStepsInEpisode) {
reward = env.act(actions[rand()%numActions]);
cumulativeReward += reward;
step++;
}
printf("Episode %d, Cumulative Reward: %d\n", episode + 1, cumulativeReward);
totalReward += cumulativeReward;
cumulativeReward = 0;
env.reset_game(); //Start the game again when the episode is over
}
return double(totalReward)/numEpisodesToEval;
}