用python/numpy固定精度的小数的单速编码
在Python中，我想获得精确的十进制数字的单速编码，直到10^{ - 3}。鉴于我的输入分数是，例如，Interval = [0.433,0.223,0.111]，它将为每个数字在间隔中生成一堆单式向量。因此，对于第一个浮点数0.433，我们应该获得4 ---＆gt; 0 0 0 0 1 0 0 0 0 0，3 ---＆gt; 0 0 0 1 0 0 0 0 0 0 0，3 ---＆gt; 0 0 0 1 0 0 0 0 0 0 0，然后将三个维度单速阵列的三个加成。
另外，我想知道这一点：即使输入编号是[0.4,0.2,0.1]，是否有一种方法可以应用与以前相同的技术？就像考虑数学上等效的数字[0.400,0.200,0.100]？
编辑：
这是我提出的尝试。我不确定这是实现结果的最佳方法，ADN此外，这不是解决我们得到的情况，例如[0.4，0.2]，但我们希望将其解释为> [0.400，0.200]。
def encode_digits(floats: list):
decimals = []
for f in floats:
decimals.append(str(f).split('0.',1)[1])
one_hot = []
for i in range(2):
for j in range(3):
temp =np.zeros(10)
temp[int(decimals[i][j])] += 1
one_hot.append(temp)
return np.array(one_hot).reshape(-1)
In Python, I would like to achieve a one-hot encoding of decimal numbers with precision until 10^{-3}. Given that my input fractions are, e.g., interval = [0.433,0.223,0.111], it would produce a stack of one-hot vectors for each number in interval. So for the first float number 0.433 we should get 4 ---> 0 0 0 0 1 0 0 0 0 0 , 3 ---> 0 0 0 1 0 0 0 0 0 0, 3 ---> 0 0 0 1 0 0 0 0 0 0 and then concatenate the three obtaining a 30 dimensional one-hot array.
Also, I'm wondering about this: even if the input numbers are, let's say, [0.4,0.2,0.1], is there a way to apply the same technique as before? Like considering mathematically equivalent numbers [0.400,0.200,0.100]?
EDIT:
This is my proposed attempt. I'm not sure this is the best way to accomplish the result, adn additionally this not solves the case where we are given for example [0.4, 0.2] but we want to interpret it as [0.400, 0.200] .
def encode_digits(floats: list):
decimals = []
for f in floats:
decimals.append(str(f).split('0.',1)[1])
one_hot = []
for i in range(2):
for j in range(3):
temp =np.zeros(10)
temp[int(decimals[i][j])] += 1
one_hot.append(temp)
return np.array(one_hot).reshape(-1)
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
如评论中所建议的，您可以乘以1000并转换为整数。此后，您可以从数字
i中提取每个单独的数字，最后应用标准单速编码：输出：
如果您希望将其连接到30维数组中，则可以做：
输出：输出：
As suggested in the comment you can multiply by 1000 and convert to integer. Thereafter you can extract each individual digit from the numbers
iand finally apply standard one-hot encoding:output:
If you prefer to concatenate it into a 30 dimensional array you could do:
output: