Python car_and_reward 예제들

예제 #1

파일: jcr_valueiter.py 프로젝트: KDmike123/Reinforcement-Learning-An-introduction

def generate_policy():
	policy_stable = True
	for i in xrange(max_car+1):
		for j in xrange(max_car+1):
			rewardlist = np.zeros(11)
			index = 0
			for k in xrange(-max_move, max_move+1):
				rewardlist[index] = car_and_reward(i, j, k, V)
				index += 1
			pi[i, j] = np.argmax(rewardlist)-5

예제 #2

파일: jack_car.py 프로젝트: zbn123/Reinforcement-Learning-An-introduction

def policy_eval():
    while True:
        delta = 0
        for i in xrange(max_car + 1):
            for j in xrange(max_car + 1):
                temp = V[i, j]
                V[i, j] = car_and_reward(i, j, pi[i, j], V)
                delta = max(delta, abs(temp - V[i, j]))
        if delta < 1e-9:
            break

예제 #3

파일: jack_car.py 프로젝트: KDmike123/Reinforcement-Learning-An-introduction

def policy_eval():
	while True:
		delta = 0
		for i in xrange(max_car+1):
			for j in xrange(max_car+1):
				temp = V[i, j]
				V[i, j] = car_and_reward(i, j, pi[i, j], V)
				delta = max(delta, abs(temp-V[i, j]))
		if delta < 1e-9:
			break

예제 #4

def generate_policy():
    policy_stable = True
    for i in xrange(max_car + 1):
        for j in xrange(max_car + 1):
            rewardlist = np.zeros(11)
            index = 0
            for k in xrange(-max_move, max_move + 1):
                rewardlist[index] = car_and_reward(i, j, k, V)
                index += 1
            pi[i, j] = np.argmax(rewardlist) - 5

예제 #5

파일: jack_car.py 프로젝트: zbn123/Reinforcement-Learning-An-introduction

def policy_improv():
    policy_stable = True
    for i in xrange(max_car + 1):
        for j in xrange(max_car + 1):
            temp = pi[i, j]
            rewardlist = np.zeros(11)
            index = 0
            for k in xrange(-max_move, max_move + 1):
                rewardlist[index] = car_and_reward(i, j, k, V)
                index += 1
            pi[i, j] = np.argmax(rewardlist) - 5
            if temp != pi[i, j]:
                policy_stable = False
    return policy_stable

예제 #6

파일: jack_car.py 프로젝트: KDmike123/Reinforcement-Learning-An-introduction

def policy_improv():
	policy_stable = True
	for i in xrange(max_car+1):
		for j in xrange(max_car+1):
			temp = pi[i, j]
			rewardlist = np.zeros(11)
			index = 0
			for k in xrange(-max_move, max_move+1):
				rewardlist[index] = car_and_reward(i, j, k, V)
				index += 1
			pi[i, j] = np.argmax(rewardlist)-5
			if temp != pi[i, j]:
				policy_stable = False
	return policy_stable

예제 #7

파일: jcr_valueiter.py 프로젝트: KDmike123/Reinforcement-Learning-An-introduction

def policy_iter():
	while True:
		delta = 0
		for i in xrange(max_car+1):
			for j in xrange(max_car+1):
				# print i,j
				temp = V[i, j]
				rewardlist = np.zeros(11)
				index = 0
				for k in xrange(-max_move, max_move+1):
					rewardlist[index] = car_and_reward(i, j, k, V)
					index += 1
				V[i, j] = np.amax(rewardlist)
				delta = max(delta, abs(temp-V[i, j]))
		if delta < 1e-9:
			break

예제 #8

def policy_iter():
    while True:
        delta = 0
        for i in xrange(max_car + 1):
            for j in xrange(max_car + 1):
                # print i,j
                temp = V[i, j]
                rewardlist = np.zeros(11)
                index = 0
                for k in xrange(-max_move, max_move + 1):
                    rewardlist[index] = car_and_reward(i, j, k, V)
                    index += 1
                V[i, j] = np.amax(rewardlist)
                delta = max(delta, abs(temp - V[i, j]))
        if delta < 1e-9:
            break