void actionOutcome(pair<State*, Action*> performed, pair<State*, double> outcome) {
State *startState = performed.first;
Action *actionPerformed = performed.second;
State *resultState = outcome.first;
double reward = outcome.second;
assert(haveSeenState(startState));
if (!haveSeenState(resultState)) {
handleNewState(resultState);
}
ActionValue *curValue = findActionValue(startState, actionPerformed);
assert(curValue != nullptr);
double newQ = reward + futureDiscount * maxQ(resultState);
curValue->value += learnRate * (newQ - curValue->value);
}
void ActionOutcome(pair<State *, Action *> performed, pair<State *, double> outcome) {
State *startState = performed.first;
Action *actionPerformed = performed.second;
State *resultState = outcome.first;
double reward = outcome.second;
if (!haveSeenState(startState)) {
handleNewState(startState);
}
if (!haveSeenState(resultState)) {
handleNewState(resultState);
}
ActionValue *curValue = findActionValue(startState, actionPerformed);
assert(curValue != nullptr);
double newQ = reward + futureDiscount * maxQ(resultState);
// cout << "r: " << reward << " fd: " << futureDiscount << " newQ: " << newQ << endl;
curValue->value += learnRate * (newQ - curValue->value);
// cout << "v : " << curValue->value << endl;
// getchar();
}