用神经网络玩FizzBuzz小游戏

来源：互联网编辑：wan玩得好手游小编更新：2024-03-19 18:28:28 人气:

csdn：https://blog.csdn.net/abcgkj

github：https://github.com/aimi-cn/AILearners

FizzBuzz是一个简单的小游戏。游戏规则如下：从1开始往上数数，当遇到3的倍数的时候，说fizz，当遇到5的倍数，说buzz，当遇到15的倍数，就说FizzBuzz，其他情况下则正常数数。

首先，我们可以写一个普通的python代码来完成这个小游戏：

def fizz_buzz_encoder(i):    if   i % 15 == 0: return 3    elif i % 5  == 0: return 2    elif i % 3  == 0: return 1    else: return 0
def fizz_buzz_decoder(i, prediction):    return [str(i), "fizz", "buzz", "fizzbuzz"][prediction]   for i in range(1,16):     print(fizz_buzz_decoder(i, fizz_buzz_encoder(i)))

output1:

12fizz4buzzfizz78fizzbuzz11fizz1314fizzbuzz

接下来，我们尝试使用PyTorch构建一个简单的神经网络模型来玩这个小游戏。

1.准备训练数据

import numpy as npimport torch
NUM_DIGITS = 10
# 用二进制numpy数组表示输入数字def binary_encode(i, num_digits):    return np.array([i >> d & 1 for d in range(num_digits)])
trX = torch.Tensor([binary_encode(i, NUM_DIGITS) for i in range(101, 2 ** NUM_DIGITS)])trY = torch.LongTensor([fizz_buzz_encoder(i) for i in range(101, 2 ** NUM_DIGITS)]) # 由于我们是4分类问题，所以Y需要为整型数据

2.用PyTorch定义模型

# 隐藏层的维度设置为100NUM_HIDDEN = 100
model = torch.nn.Sequential(    torch.nn.Linear(NUM_DIGITS, NUM_HIDDEN),    torch.nn.ReLU(),    torch.nn.Linear(NUM_HIDDEN, 4))
# 使用cuda加速if torch.cuda.is_available():    model = model.cuda()

3.定义损失函数和优化函数

# 由于是分类问题，我们选择交叉熵函数loss_fn = torch.nn.CrossEntropyLoss()optimizer = torch.optim.SGD(model.parameters(), lr = 0.05)

4.开始训练

BATCH_SIZE = 128for epoch in range(10000):    for start in range(0, len(trX), BATCH_SIZE):        end = start + BATCH_SIZE        batchX = trX[start:end]        batchY = trY[start:end]        # 将数据放到cuda上        if torch.cuda.is_available():            batchX = batchX.cuda()            batchY = batchY.cuda()        y_pred = model(batchX)        loss = loss_fn(y_pred, batchY)
        optimizer.zero_grad()        loss.backward()        optimizer.step()
    # 打印loss    print('Epoch:', epoch, 'Loss:', loss.item())

5.准备测试数据

# 我们用训练好的模型尝试在1到100这些数字上玩FizzBuzz游戏testX = torch.Tensor([binary_encode(i, NUM_DIGITS) for i in range(1, 101)])
if torch.cuda.is_available():    testX = testX.cuda()

6.测试

# 由于测试时不需要保存梯度，故把测试时的梯度清空with torch.no_grad():    testY = model(testX)predictions = zip(range(1, 101), list(testY.max(1)[1].data.tolist()))print([fizz_buzz_decoder(i, x) for (i, x) in predictions])
testY = testY.cpu()print("accuracy:", np.sum(testY.max(1)[1].numpy() == np.array([fizz_buzz_encoder(i) for i in range(1,101)])), "%")

output2：

['1', 'buzz', 'fizz', '4', 'buzz', 'fizz', '7', '8', 'fizz', 'buzz', '11', 'fizz', '13', '14', 'fizzbuzz', '16', '17', 'fizz', 'fizz', 'buzz', 'fizz', '22', '23', 'fizz', 'buzz', '26', 'fizz', '28', '29', 'fizzbuzz', '31', 'buzz', 'fizz', '34', 'buzz', 'fizz', '37', '38', 'fizz', 'buzz', '41', 'fizz', '43', '44', 'fizzbuzz', '46', '47', 'fizz', '49', 'buzz', 'fizz', '52', '53', 'fizz', 'buzz', '56', 'fizz', '58', '59', 'fizzbuzz', '61', '62', 'fizz', '64', 'buzz', 'fizz', '67', '68', 'fizz', 'buzz', '71', 'fizz', '73', '74', 'fizzbuzz', '76', '77', 'fizz', '79', 'buzz', 'fizz', '82', 'fizz', 'fizz', 'buzz', '86', 'fizz', '88', '89', 'fizzbuzz', '91', '92', 'fizz', '94', 'buzz', 'fizz', '97', '98', 'fizz', 'buzz']accuracy: 96 %