我有一个稀疏矢量列表:
print(type(downsample_matrix)) # Display <class 'list'>
print(type(downsample_matrix[0])) # Display <class 'scipy.sparse.csr.csr_matrix'>
我想在cosine_similarity
上使用功能scikit learn downsampled_matrix
但是我收到以下错误:
ValueError Traceback (most recent call last)
<ipython-input-27-5997ca6abb2d> in <module>()
19 downsample_matrix.append(vector)
20 downsample_coefficient = 0
---> 21 similarity_matrix = cosine_similarity(downsample_matrix)
22 plt.matshow(similarity_matrix)
23 plt.show()
/home/venv/lib/python3.5/site-packages/sklearn/metrics/pairwise.py in cosine_similarity(X, Y, dense_output)
908 # to avoid recursive import
909
--> 910 X, Y = check_pairwise_arrays(X, Y)
911
912 X_normalized = normalize(X, copy=True)
/home/venv/lib/python3.5/site-packages/sklearn/metrics/pairwise.py in check_pairwise_arrays(X, Y, precomputed, dtype)
104 if Y is X or Y is None:
105 X = Y = check_array(X, accept_sparse='csr', dtype=dtype,
--> 106 warn_on_dtype=warn_on_dtype, estimator=estimator)
107 else:
108 X = check_array(X, accept_sparse='csr', dtype=dtype,
/home/venv/lib/python3.5/site-packages/sklearn/utils/validation.py in check_array(array, accept_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, warn_on_dtype, estimator)
380 force_all_finite)
381 else:
--> 382 array = np.array(array, dtype=dtype, order=order, copy=copy)
383
384 if ensure_2d:
ValueError: setting an array element with a sequence.
当我的列表由nd.array
组成时,我没有问题:
print(type(downsample_matrix)) # Display <class 'list'>
print(type(downsample_matrix[0])) # Display <class 'numpy.ndarray'>
如何在我的sparce向量列表中应用cosine_similarity?
答案 0 :(得分:2)
创建一个小的稀疏矩阵。请注意,它不是ndarray
的子类。它将数据存储在3个数组中 - 数据和索引:
In [196]: M = sparse.csr_matrix([[0,1,0],[1,0,1]])
In [197]: M
Out[197]:
<2x3 sparse matrix of type '<class 'numpy.int32'>'
with 3 stored elements in Compressed Sparse Row format>
In [198]: M.data
Out[198]: array([1, 1, 1], dtype=int32)
In [199]: M.indices
Out[199]: array([1, 0, 2], dtype=int32)
In [200]: M.indptr
Out[200]: array([0, 1, 3], dtype=int32)
如果我尝试从这个矩阵的列表中创建一个数组,我得到一个对象dtype数组,有3个元素(指向这个矩阵的指针):
In [201]: alist = [M,M,M]
In [202]: np.array(alist)
Out[202]: /usr/local/lib/python3.5/dist-packages/scipy/sparse/compressed.py:294: SparseEfficiencyWarning: Comparing sparse matrices using >= and <= is inefficient, using <, >, or !=, instead.
"using <, >, or !=, instead.", SparseEfficiencyWarning)
array([ <2x3 sparse matrix of type '<class 'numpy.int32'>'
with 3 stored elements in Compressed Sparse Row format>,
<2x3 sparse matrix of type '<class 'numpy.int32'>'
with 3 stored elements in Compressed Sparse Row format>,
<2x3 sparse matrix of type '<class 'numpy.int32'>'
with 3 stored elements in Compressed Sparse Row format>], dtype=object)
如果另外我指定了dtype,我会收到你的错误:
In [203]: np.array(alist,dtype=int)
...
ValueError: setting an array element with a sequence.
它无法将列表转换为数字数组。
但如果它是密集数组的列表,我得到一个3d数组:
In [204]: np.array([M.A,M.A,M.A],dtype=int)
Out[204]:
array([[[0, 1, 0],
[1, 0, 1]],
[[0, 1, 0],
[1, 0, 1]],
[[0, 1, 0],
[1, 0, 1]]])
In [205]: _.shape
Out[205]: (3, 2, 3)
我还可以使用稀疏版vstack
或hstack
连接稀疏矩阵。
In [206]: sparse.vstack(alist)
Out[206]:
<6x3 sparse matrix of type '<class 'numpy.int32'>'
with 9 stored elements in Compressed Sparse Row format>
In [207]: _.A
Out[207]:
array([[0, 1, 0],
[1, 0, 1],
[0, 1, 0],
[1, 0, 1],
[0, 1, 0],
[1, 0, 1]], dtype=int32)
注意形状,(6,3)。稀疏矩阵总是2d。
sparse.vstack
将任务传递给sparse.bmat
,coo
从'blocks'构造一个新的稀疏矩阵。它是通过将块的cosine_similarity
表示与适当的偏移相结合来实现的。
由于sparse.vstack
需要2d数组或稀疏矩阵,因此您必须使用In [212]: cosine_similarity(sparse.vstack(alist))
Out[212]:
array([[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.],
[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.],
[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.]])
In [213]: cosine_similarity( np.array([M.A,M.A,M.A],dtype=int).reshape(-1,3))
Out[213]:
array([[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.],
[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.],
[ 1., 0., 1., 0., 1., 0.],
[ 0., 1., 0., 1., 0., 1.]])
来连接矩阵。或者重塑3d阵列连接的结果
Car
int id PK
int modelId FK
CarDetails
int carId PK, FK to Car.id
varchar(50) description
答案 1 :(得分:0)
试试这个
pairwise dense output:
[[ 1. 0.36514837 0.31622777]
[ 0.36514837 1. 0.28867513]
[ 0.31622777 0.28867513 1. ]]
结果
print (scipy.__version__)
0.19.0
我的狡猾
{{1}}