blocks|key|4410274|text|None是NP.newaxis的别名。它创建一个长度为1的轴。这对于矩阵乘法等很有用。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|4410275|>>>>+import+numpy+as+NP
>>>>+a+=+NP.arange(1,5)
>>>>+print+a
[1+2+3+4]
>>>>+print+a.shape
(4,)
>>>>+print+a[:,None].shape
(4,+1)
>>>>+print+a[:,None]
[[1]
+[2]
+[3]
+[4]]++++|code-block|syntax|javascript|4410276|entityMap|0|LINK|mutability|MUTABLE|url|http://docs.scipy.org/doc/numpy/reference/arrays.indexing.html#numpy.newaxis^0|0|4|5|A|0|0|0^^$0|@$1|2|3|4|5|6|7|S|8|@$9|T|A|U|B|C]]|D|@$9|V|A|W|1|X]]|E|$]]|$1|F|3|G|5|H|7|Y|8|@]|D|@]|E|$I|J]]|$1|K|3|-4|5|6|7|Z|8|@]|D|@]|E|$]]]|L|$M|$5|N|O|P|E|$Q|R]]]]

<code>None</code> is an alias for <a href="http://docs.scipy.org/doc/numpy/reference/arrays.indexing.html#numpy.newaxis" rel="noreferrer">NP.newaxis</a>. It creates an axis with length 1. This can be useful for matrix multiplcation etc.

<pre><code>&gt;&gt;&gt;&gt; import numpy as NP
&gt;&gt;&gt;&gt; a = NP.arange(1,5)
&gt;&gt;&gt;&gt; print a
[1 2 3 4]
&gt;&gt;&gt;&gt; print a.shape
(4,)
&gt;&gt;&gt;&gt; print a[:,None].shape
(4, 1)
&gt;&gt;&gt;&gt; print a[:,None]
[[1]
 [2]
 [3]
 [4]] 
</code></pre>

blocks|key|4410299|text|我来这里之前，在做同样的Udacity课程时遇到了完全相同的问题。我想做的是转置一维Numpy序列/数组，它不适用于numpy.transpose(1，2，3)。所以我想添加你可以像这样转置(source)：|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|4410300|numpy.matrix([1,+2,+3]).T|code-block|syntax|javascript|4410301|其结果是：|4410302|matrix([[1],
++++++++[2],
++++++++[3]])|4410303|它与几乎相同(类型不同)：|4410304|x=np.array([1,+2,+3])
x[:,None]|4410305|但我觉得这样更容易记住。|4410306|entityMap|0|LINK|mutability|MUTABLE|url|https://stackoverflow.com/a/11885581/828184^0|2O|6|0|0|0|0|0|0|0|0^^$0|@$1|2|3|4|5|6|7|10|8|@]|9|@$A|11|B|12|1|13]]|C|$]]|$1|D|3|E|5|F|7|14|8|@]|9|@]|C|$G|H]]|$1|I|3|J|5|6|7|15|8|@]|9|@]|C|$]]|$1|K|3|L|5|F|7|16|8|@]|9|@]|C|$G|H]]|$1|M|3|N|5|6|7|17|8|@]|9|@]|C|$]]|$1|O|3|P|5|F|7|18|8|@]|9|@]|C|$G|H]]|$1|Q|3|R|5|6|7|19|8|@]|9|@]|C|$]]|$1|S|3|-4|5|6|7|1A|8|@]|9|@]|C|$]]]|T|$U|$5|V|W|X|C|$Y|Z]]]]

I came here after having the exact same problem doing the same Udacity course. What I wanted to do is transpose the one dimensional numpy series/array which does not work with numpy.transpose([1, 2, 3]).
So I wanted to add you can transpose like this (<a href="https://stackoverflow.com/a/11885581/828184">source</a>): 

<pre><code>numpy.matrix([1, 2, 3]).T
</code></pre>

It results in:

<pre><code>matrix([[1],
 [2],
 [3]])
</code></pre>

which is pretty much identical (type is different) to:

<pre><code>x=np.array([1, 2, 3])
x[:,None]
</code></pre>

But I think it's easier to remember...

blocks|key|4410322|text|简单地说，它允许在不同维数的两个数组之间进行操作。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4410323|它通过添加一个新的空维度来实现这一点，该维度将自动适应另一个数组的大小。|4410324|所以基本上如果：|4410325|Array1+=+shape100和Array2+=+shape10,100|4410326|Array1+*+Array2通常会给出一个错误。|offset|length|style|CODE|4410327|Array1[:,None]+*+Array2将起作用。|4410328|entityMap^0|0|0|0|0|0|F|0|0|N|0^^$0|@$1|2|3|4|5|6|7|R|8|@]|9|@]|A|$]]|$1|B|3|C|5|6|7|S|8|@]|9|@]|A|$]]|$1|D|3|E|5|6|7|T|8|@]|9|@]|A|$]]|$1|F|3|G|5|6|7|U|8|@]|9|@]|A|$]]|$1|H|3|I|5|6|7|V|8|@$J|W|K|X|L|M]]|9|@]|A|$]]|$1|N|3|O|5|6|7|Y|8|@$J|Z|K|10|L|M]]|9|@]|A|$]]|$1|P|3|-4|5|6|7|11|8|@]|9|@]|A|$]]]|Q|$]]

to explain it in plain english, it allows operations between two arrays of different number of dimensions.

It does this by adding a new, empty dimension which will automagically fit the size of the other array.

So basically if:

Array1 = shape[100]
and 
 Array2 = shape[10,100]

<code>Array1 * Array2</code> will normally give an error.

<code>Array1[:,None] * Array2</code> will work.

blocks|key|4410343|text|如果您看到有经验的NumPy用户编写的代码，您通常会看到他们使用特殊的分片语法，而不是调用重塑。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4410344|x+=+v[None,+:]|code-block|syntax|javascript|4410345|或|4410346|x+=+v[:,+None]|4410347|这些行创建了一个切片，它查看v的所有项，但要求NumPy为关联的轴添加大小为1的新维度。|4410348|entityMap^0|0|0|0|0|0^^$0|@$1|2|3|4|5|6|7|O|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|P|8|@]|9|@]|A|$E|F]]|$1|G|3|H|5|6|7|Q|8|@]|9|@]|A|$]]|$1|I|3|J|5|D|7|R|8|@]|9|@]|A|$E|F]]|$1|K|3|L|5|6|7|S|8|@]|9|@]|A|$]]|$1|M|3|-4|5|6|7|T|8|@]|9|@]|A|$]]]|N|$]]

If you see code from experienced NumPy users, you will often see them use a special slicing syntax instead of calling reshape. 

<pre><code>x = v[None, :]
</code></pre>

or

<pre><code>x = v[:, None]
</code></pre>

Those lines create a slice that looks at all of the items of v but asks NumPy to add a new dimension of size 1 for the associated axis.

blocks|key|4410358|text|(为了澄清@GWW的答案和@BlueRine+S的评论)在使用numpy数组时，将一维数组清晰地视为行或列向量是一个好主意。Andrew+Ng也指出了这一点，以避免代码中的bug。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4410359|>>>+import+numpy+as+NP
>>>+a+=+NP.arange(1,5)
>>>+a
array([1,+2,+3,+4])
>>>+a.shape
(4,)
>>>+a[:,None].shape
(4,+1)
>>>+a[:,None]
array([[1],
+++++++[2],
+++++++[3],
+++++++[4]])
>>>+a[None,:].shape
(1,+4)
>>>+a[None,:]
array([[1,+2,+3,+4]])
>>>+np.reshape(a,+(1,+-1))
array([[1,+2,+3,+4]])
>>>+np.reshape(a,+(-1,+1))
array([[1],
+++++++[2],
+++++++[3],
+++++++[4]])|code-block|syntax|javascript|4410360|entityMap^0|0|0^^$0|@$1|2|3|4|5|6|7|I|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|J|8|@]|9|@]|A|$E|F]]|$1|G|3|-4|5|6|7|K|8|@]|9|@]|A|$]]]|H|$]]

(To clarify the answer by @GWW and the comment by @BlueRine S) While working with numpy arrays it is a good idea to clearly treat one dimensional arrays as row or column vectors. This has been pointed out by Andrew Ng also, to avoid bugs in the code.
<pre><code>&gt;&gt;&gt; import numpy as NP
&gt;&gt;&gt; a = NP.arange(1,5)
&gt;&gt;&gt; a
array([1, 2, 3, 4])
&gt;&gt;&gt; a.shape
(4,)
&gt;&gt;&gt; a[:,None].shape
(4, 1)
&gt;&gt;&gt; a[:,None]
array([[1],
 [2],
 [3],
 [4]])
&gt;&gt;&gt; a[None,:].shape
(1, 4)
&gt;&gt;&gt; a[None,:]
array([[1, 2, 3, 4]])
&gt;&gt;&gt; np.reshape(a, (1, -1))
array([[1, 2, 3, 4]])
&gt;&gt;&gt; np.reshape(a, (-1, 1))
array([[1],
 [2],
 [3],
 [4]])
</code></pre>

I'm taking the Udacity course on deep learning and I came across the following code:

<pre><code>def reformat(dataset, labels):
 dataset = dataset.reshape((-1, image_size * image_size)).astype(np.float32)
 # Map 0 to [1.0, 0.0, 0.0 ...], 1 to [0.0, 1.0, 0.0 ...]
 labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32)
 return dataset, labels
</code></pre>

What does <code>labels[:,None]</code> actually do here?

In numpy, what does selection by [:,None] do?

我正在学习关于深度学习的Udacity课程，我遇到了以下代码：def reformat(dataset, labels):    dataset = dataset.reshape((-1, image_size * image_size)).astype(np.float32)    # Map 0 to [1.0,...

问在numpy中，selection by [:，None]做了什么？
EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问在numpy中，selection by [:，None]做了什么？EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问在numpy中，selection by [:，None]做了什么？
EN