Save and load numpy matrices and vectors, and use the saved vectors for similarity calculation

numpySaving and loading of matrices and vectors,And use the saved vector for similarity calculation

import numpy as np
from sklearn.metrics.pairwise import cosine_similarity

使用 arange创建array数组

a = np.arange(0, 5)
print(a)  # [0 1 2 3 4]
b = np.arange(5, 15)
print(b)  # [ 5 6 7 8 9 10 11 12 13 14]
b = b.reshape(2, 5)
print(b)
""" [[ 5 6 7 8 9] [10 11 12 13 14]] """

使用列表创建ndarray

li = [1, 2, 3, 4]
li1 = [[1, 2, 3, 5], [4, 5, 6, 8]]

li_arr = np.array(li)  # # ndarry
print(li_arr)  # [1 2 3 4]

li1_arr = np.array(li1)  # ndarry
print(li1_arr)
""" [[1 2 3 5] [4 5 6 8]] """

Save a matrix or vector

# 保存方式1 保存为npy
np.save('arr.npy', li1_arr)
# 保存方式2 保存为npz npzMultiple matrix vectors can be saved 其实就是将多个npyThe packaged object is saved Default is object 键从attr1开始
np.savez('arr.npz', li1_arr, li_arr)

Load and save a matrix or vector

npy_res = np.load('arr.npy')
print(npy_res)
""" [[1 2 3 5] [4 5 6 8]] """

Calculate similarity using the read vector

def cosine_similarity_(a, b):
    """ Find cosine similarity or distance :param a: :param b: :return: """
    res = cosine_similarity(a.reshape(1, -1), b.reshape(1, -1))
    return res[0][0]


cal_arr = np.asarray([1, 3, 4, 5])
print(cal_arr)  # [1 3 4 5]

""" 计算npy_res中的每个向量与cal_arr的相似度 """
for arr in npy_res:
    print(cosine_similarity_(arr, cal_arr))  # 0.9865867645279247 0.9787763071201612

""" 计算npz中的arr1A vector AND in an objectcal_arr的相似度 """
for arr in npz_res['arr_0']:
    print(arr)
    print(cosine_similarity_(arr, cal_arr))  # 0.9865867645279247 0.9787763071201612

完整代码

import numpy as np
from sklearn.metrics.pairwise import cosine_similarity

""" 使用 arange创建array数组 """
a = np.arange(0, 5)
print(a)  # [0 1 2 3 4]
b = np.arange(5, 15)
print(b)  # [ 5 6 7 8 9 10 11 12 13 14]
b = b.reshape(2, 5)
print(b)
""" [[ 5 6 7 8 9] [10 11 12 13 14]] """

""" 使用列表创建ndarray """
li = [1, 2, 3, 4]
li1 = [[1, 2, 3, 5], [4, 5, 6, 8]]

li_arr = np.array(li)  # # ndarry
print(li_arr)  # [1 2 3 4]

li1_arr = np.array(li1)  # ndarry
print(li1_arr)
""" [[1 2 3 5] [4 5 6 8]] """

# Save a matrix or vector
# 保存方式1 保存为npy
np.save('arr.npy', li1_arr)
# 保存方式2 保存为npz npzMultiple matrix vectors can be saved 其实就是将多个npyThe packaged object is saved Default is object 键从attr1开始
np.savez('arr.npz', li1_arr, li_arr)

# 加载保存的npy文件
npy_res = np.load('arr.npy')
print(npy_res)
""" [[1 2 3 5] [4 5 6 8]] """

# 加载保存的npz文件
npz_res = np.load('arr.npz')
npz_res = dict(npz_res)
print(npz_res)
""" {'arr_0': array([[1, 2, 3, 5], [4, 5, 6, 8]]), 'arr_1': array([1, 2, 3, 4])} """

""" Calculate the similarity of the read vectors """


def cosine_similarity_(a, b):
    """ Find cosine similarity or distance :param a: :param b: :return: """
    res = cosine_similarity(a.reshape(1, -1), b.reshape(1, -1))
    return res[0][0]


cal_arr = np.asarray([1, 3, 4, 5])
print(cal_arr)  # [1 3 4 5]

""" 计算npy_res中的每个向量与cal_arr的相似度 """
for arr in npy_res:
    print(cosine_similarity_(arr, cal_arr))  # 0.9865867645279247 0.9787763071201612

""" 计算npz中的arr1A vector AND in an objectcal_arr的相似度 """
for arr in npz_res['arr_0']:
    print(arr)
    print(cosine_similarity_(arr, cal_arr))  # 0.9865867645279247 0.9787763071201612