python – Numpy-从2D数组获取邻居矩阵

发布时间：2020-12-20 11:47:03 所属栏目：Python 来源：网络整理

导读：我有一个无向图,由下式给出： A,BA,CA,FB,DC,FE,DE,F 我需要将其转换为矩阵N,其中 N[x,y] = 1 if x is neighbor of y (like A and B) and 0 if not 最快的方法是什么？ NOTA图表存储为numpy字符串数组： array([['A','B'],['A','C'],'F'],['B','D'],['C',['E

我有一个无向图,由下式给出：

A,B
A,C
A,F
B,D
C,F
E,D
E,F

我需要将其转换为矩阵N,其中

N[x,y] = 1 if x is neighbor of y (like A and B) and 0 if not

最快的方法是什么？

NOTA图表存储为numpy字符串数组：

array([['A','B'],['A','C'],'F'],['B','D'],['C',['E','F']],dtype='|S4')

预期产量：

A B C D E F
A  1 1 1 0 0 1
B  1 1 0 1 0 0 
C  1 0 1 1 0 1
D  0 1 1 1 1 0
E  0 0 0 1 1 1 
F  1 0 1 0 1 1

谢谢

解决方法

这是一种NumPythonic方法 –

# Tag each string with a numeric ID based on the uniqueness among other strings
_,ID = np.unique(graph,return_inverse=True)
M = ID.reshape(graph.shape)

# Consider each row of numeric IDs as an indexing tuple of a 2D array.
# Calculate the linear indices corresponding to each tuple.
n = M.max()+1
idx1 = M[:,0]*(n) + M[:,1]
idx2 = M[:,1]*(n) + M[:,0]

# Setup output array with 1s on the diagonal.
out = np.eye(n,dtype=int)

# Put 1s at places specified by the earlier computed pairs of linear indices
np.put(out,[idx1,idx2],1)

样本输入,输出 –

In [93]: graph
Out[93]: 
array([['A',dtype='|S4')

In [94]: out
Out[94]: 
array([[1,1,1],[1,0],[0,1]])

运行时测试

本节基本上比较了迄今为止列出的所有方法(在这篇文章和其他帖子中)解决案例的表现.

定义功能 –

def networkx_based(a): #@atomh33ls's solution code
    A=nx.Graph()
    for x in a:
        A.add_edge(x[0],x[1])
    return nx.adjacency_matrix(A)

def vectorize_based(data): # @plonser's solution code
    ord_u = np.vectorize(lambda x: ord(x)-65)
    s = ord_u(data)
    res = np.zeros((s.max()+1,s.max()+1),dtype=int)
    res[s[:,s[:,1]] = 1
    res_sym = res + res.T
    np.fill_diagonal(res_sym,1)
    return res_sym

def unique_based(graph): #solution from this post
    _,return_inverse=True)
    M = ID.reshape(graph.shape)
    n = M.max()+1
    idx1 = M[:,1]
    idx2 = M[:,0]
    out = np.eye(n,dtype=int)
    np.put(out,1)
    return out

计时 –

In [321]: N = 1000

In [322]: arr = np.random.randint(65,90,(N,2))

In [323]: graph = np.array([map(chr,a) for a in arr],dtype='S4')

In [324]: %timeit networkx_based(graph)
100 loops,best of 3: 3.79 ms per loop

In [325]: %timeit vectorize_based(graph)
1000 loops,best of 3: 760 μs per loop

In [326]: %timeit unique_based(graph)
1000 loops,best of 3: 405 μs per loop

（编辑：李大同）

【声明】本站内容均来自网络，其相关言论仅代表作者个人观点，不代表本站立场。若无意侵犯到您的权利，请及时与联系站长删除相关内容!