當前位置：首頁 > 编程资源 > 编程问答 >内容正文

编程问答

numpy---(精简)

發(fā)布時間：2024/1/17 编程问答 20 豆豆

生活随笔收集整理的這篇文章主要介紹了 numpy---(精简) 小編覺得挺不錯的,現(xiàn)在分享給大家,幫大家做個參考.

numpy get started

導入numpy庫, 并查看版本

import numpy as np np.__version__ '1.14.3' # pyplot顯示畫圖, 數(shù)據(jù)分析與可視化 import matplotlib.pyplot as plt lena = plt.imread('lena.jpg') # type(lena) lena array([[[225, 138, 128],[224, 137, 127],[223, 136, 126],...,[234, 146, 126],[220, 129, 110],[197, 104, 86]],[[222, 138, 127],[224, 137, 127],[224, 137, 127],...,[235, 150, 130],[218, 131, 112],[189, 102, 83]],[[222, 138, 127],[224, 137, 127],[225, 138, 128],...,[230, 148, 127],[215, 133, 112],[190, 105, 85]],...,[[ 82, 21, 55],[ 81, 20, 54],[ 92, 28, 62],...,[175, 71, 82],[175, 68, 78],[175, 65, 74]],[[ 80, 18, 55],[ 81, 20, 54],[ 94, 33, 67],...,[177, 69, 82],[182, 70, 82],[183, 72, 81]],[[ 81, 19, 56],[ 83, 21, 58],[ 96, 35, 69],...,[178, 68, 81],[183, 71, 83],[188, 74, 84]]], dtype=uint8) lena2 = lena - 10 plt.imshow(lena2) plt.show()

創(chuàng)建ndarray

使用np.array()由python list創(chuàng)建

numpy默認ndarray的所有元素的類型是相同的
如果傳遞的列表中包含不同的類型, 則統(tǒng)一為同一類型, 優(yōu)先級:str > float > int

n1 = np.array([3, 1, 4, 5]) n1 array([3, 1, 4, 5]) n2 = np.array([[2, 3, 4, 5], [4, 6, 1, 9], [5, 6, 7, 8]]) n2 array([[2, 3, 4, 5],[4, 6, 1, 9],[5, 6, 7, 8]]) type(n2) # shape是屬性,不是方法 n2.shape (3, 4) n1.shape (4,) # 行, 列, 維度 # 一張二維圖片轉化成數(shù)組為三位數(shù)組 lena.shape (512, 512, 3) n3 = np.array(['ABC', 1, 3.14]) n3 array(['ABC', '1', '3.14'], dtype='<U4')

使用np.routines函數(shù)創(chuàng)建

np.ones(shape=(10, 8), dtype=int) array([[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1],[1, 1, 1, 1, 1, 1, 1, 1]]) ones = np.ones(shape=(100, 80, 3), dtype=float) plt.imshow(ones) plt.show()

np.zeros((4, 4)) array([[0., 0., 0., 0.],[0., 0., 0., 0.],[0., 0., 0., 0.],[0., 0., 0., 0.]]) np.full((10, 10), fill_value=1024) array([[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024],[1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024]]) # np.eye(N, M=None, k=0, dtype=<class 'float'>, order='C') # 對角線為1, 其他位置為0, 滿秩矩陣 np.eye(10) array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],[0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],[0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],[0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],[0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],[0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],[0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],[0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],[0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],[0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]]) # np.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None) np.linspace(0, 100, 20) array([ 0. , 5.26315789, 10.52631579, 15.78947368,21.05263158, 26.31578947, 31.57894737, 36.84210526,42.10526316, 47.36842105, 52.63157895, 57.89473684,63.15789474, 68.42105263, 73.68421053, 78.94736842,84.21052632, 89.47368421, 94.73684211, 100. ]) # np.arange([start,] stop[, step,], dtype=None) # 左閉右開 np.arange(0, 100, 5) array([ 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80,85, 90, 95]) # np.random.randint(low, high=None, size=None, dtype='l') np.random.randint(0, 150, 5) array([127, 110, 56, 63, 77]) # np.random.randn(d0, d1, ..., dn) # 標準正態(tài)分布 np.random.randn(100) array([-0.41595026, 1.47042723, 0.03288821, 0.85004019, 0.7950821 ,0.13841712, 0.28218393, 1.22280226, -0.55662926, -0.85044176,0.87454005, -1.56832096, 1.69536713, 0.12126746, 1.05180469,0.78131875, -0.70417438, -0.58430437, -0.49943889, 0.2231934 ,-0.55686039, -0.48619634, 0.48127741, 0.27138361, 0.17976988,-0.11101901, -0.71860609, 1.2858034 , 0.26501713, 0.15528386,-0.3639874 , -0.50213498, -1.3080041 , -2.35717083, -0.648195 ,0.24412035, -1.50979317, 1.09656183, 0.00946873, -0.73389828,-0.04357452, 0.80743789, -1.72143062, 0.10460993, 1.32929205,0.06736016, -1.56802382, 0.90329101, -0.45505224, 0.04915999,0.26430304, -0.40411427, 0.42802756, -1.69807546, 0.27891151,1.05498128, 0.94090423, -0.42022151, 1.65546614, -0.22287079,0.69203073, 0.96486237, -1.28087795, 0.75158138, -0.18673762,-0.70781096, -1.71156378, 0.65202125, -0.3525935 , 0.4323014 ,-0.63716862, -0.67085324, -0.30546365, 0.39392657, -2.13986037,-0.0085726 , -1.67360167, 1.84832111, 0.0671747 , 0.01600444,0.52551343, -0.60296408, -0.47100002, -2.18264449, 0.46744126,0.72398992, -1.71408793, 0.14587077, -0.18404951, -0.80683105,1.73309297, 0.35799329, 0.73527189, 0.13199485, 0.26461892,0.54344243, 0.49003007, 0.21602823, -1.22451068, -0.21714807]) # np.random.normal(loc=0.0, scale=1.0, size=None) # scale波動 np.random.normal(loc=175, scale=1, size=100) array([176.21310971, 172.20006366, 175.29247008, 173.66475082,173.68890116, 174.71321419, 175.80304124, 175.10018316,173.44016299, 174.80136342, 175.37660695, 177.19022468,176.32440094, 173.14755284, 175.06826748, 176.42310704,174.76973001, 172.07888002, 174.81805161, 175.85111712,176.1612796 , 175.9304326 , 174.08051939, 174.2482614 ,172.47992484, 174.73893155, 173.8072 , 173.57799107,173.78284387, 176.32936172, 175.17084547, 173.21273207,175.28091245, 174.47728685, 174.24929528, 174.37795464,173.35172255, 175.33469387, 174.38263904, 176.28884503,174.48028776, 176.07224738, 175.28880278, 177.13037103,171.68068476, 174.58779908, 177.3445544 , 174.96102577,173.9927033 , 174.81596921, 173.40709395, 175.09461029,174.56116781, 176.10069031, 177.34382616, 176.63857035,175.29170695, 173.77097116, 173.92263266, 177.1159495 ,175.33183934, 175.41897696, 174.01483045, 175.26064743,174.52707392, 174.71789507, 175.83135718, 175.3980088 ,175.28031481, 176.63722956, 176.14911054, 174.1617964 ,174.12355257, 175.97611042, 175.4970436 , 176.42210635,173.54120183, 174.25305399, 172.89636185, 175.76694058,172.44363816, 172.97763963, 173.76208303, 175.68367144,174.39331671, 174.26906247, 173.97178951, 174.34262788,174.78171771, 176.3154983 , 175.18898772, 175.03515302,175.01803086, 175.89679058, 174.81759265, 174.66847045,175.67714752, 173.83397302, 172.11278424, 174.91772609]) # 生成0到1的隨機數(shù), 左閉右開 # 使用隨機數(shù)生成一張圖片 r = np.random.random(size=(200, 300, 3)) plt.imshow(r) plt.show()

ndarray的屬性

4個必記參數(shù): ndim: 維度
shape: 形狀(各維度的長度)
size: 總長度
dtype: 元素類型

ndarray的基本操作

索引

# 二維數(shù)組索引 n5 = np.random.randint(0, 100, (3, 4)) n5 array([[44, 50, 39, 56],[29, 50, 49, 95],[11, 20, 97, 73]]) n5[0, 1] 50 # 三位數(shù)組索引 n6 = np.random.randint(0, 100, (3, 4, 5)) n6 array([[[83, 35, 84, 88, 18],[62, 37, 55, 65, 8],[26, 86, 50, 11, 37],[37, 93, 1, 86, 71]],[[33, 25, 72, 13, 82],[80, 36, 69, 37, 32],[43, 79, 40, 3, 46],[67, 10, 79, 98, 58]],[[44, 36, 89, 64, 86],[82, 9, 37, 33, 13],[59, 55, 45, 59, 29],[72, 68, 88, 23, 64]]]) # 可以看成一維和二維組合 n6[0, 3, 1] 93

切片

n7 = np.random.randint(150, size=10) n7 array([ 12, 144, 141, 103, 82, 119, 85, 83, 36, 45]) # 和python list一樣左閉右開 # 一維數(shù)組切片 n7[0:5] array([ 12, 144, 141, 103, 82]) n6.shape (3, 4, 5) n6 array([[[83, 35, 84, 88, 18],[62, 37, 55, 65, 8],[26, 86, 50, 11, 37],[37, 93, 1, 86, 71]],[[33, 25, 72, 13, 82],[80, 36, 69, 37, 32],[43, 79, 40, 3, 46],[67, 10, 79, 98, 58]],[[44, 36, 89, 64, 86],[82, 9, 37, 33, 13],[59, 55, 45, 59, 29],[72, 68, 88, 23, 64]]]) # 三位數(shù)組切片 n6[0:2] array([[[83, 35, 84, 88, 18],[62, 37, 55, 65, 8],[26, 86, 50, 11, 37],[37, 93, 1, 86, 71]],[[33, 25, 72, 13, 82],[80, 36, 69, 37, 32],[43, 79, 40, 3, 46],[67, 10, 79, 98, 58]]]) n6[0:2, 1:3] array([[[62, 37, 55, 65, 8],[26, 86, 50, 11, 37]],[[80, 36, 69, 37, 32],[43, 79, 40, 3, 46]]]) n6[0:2, 1:3, -2:] array([[[65, 8],[11, 37]],[[37, 32],[ 3, 46]]]) # 將數(shù)據(jù)反轉 n8 = np.arange(0, 10, 1) n8 array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]) n8[::-1] array([9, 8, 7, 6, 5, 4, 3, 2, 1, 0]) n8[::-2] array([9, 7, 5, 3, 1])

變形reshape

使用reshape函數(shù), 注意參數(shù)是一個tuple

# a.reshape(shape, order='C') n8.reshape((5, 2)) array([[0, 1],[2, 3],[4, 5],[6, 7],[8, 9]]) # 對圖片進行reshape lena.reshape(512 * 512 * 3) array([225, 138, 128, ..., 188, 74, 84], dtype=uint8) # 如果是負數(shù), 直接轉換成一維的數(shù)組ndarray lena.reshape(-1) array([225, 138, 128, ..., 188, 74, 84], dtype=uint8)

級聯(lián)

np.concatenate() 級聯(lián)需要注意的事項:
1. 級聯(lián)的參數(shù)是列表, 一定要加中括號或小括號
2. 維度必須相同
3. 形狀相符
4. 級聯(lián)的方向默認是shape元組的第一個值代表的維度方向
5. 可以通過axis參數(shù)改變級聯(lián)的方向

import numpy as np n9 = np.random.randint(0, 10, size=(5, 5)) n9 array([[1, 4, 8, 8, 1],[0, 0, 8, 4, 0],[3, 1, 2, 3, 5],[9, 6, 5, 8, 0],[3, 6, 1, 7, 2]]) # np.concatenate((a1, a2, ...), axis=0, out=None) np.concatenate((n9, n9)) array([[1, 4, 8, 8, 1],[0, 0, 8, 4, 0],[3, 1, 2, 3, 5],[9, 6, 5, 8, 0],[3, 6, 1, 7, 2],[1, 4, 8, 8, 1],[0, 0, 8, 4, 0],[3, 1, 2, 3, 5],[9, 6, 5, 8, 0],[3, 6, 1, 7, 2]]) n9.ndim 2 np.concatenate((n9, n9), axis=1) array([[1, 4, 8, 8, 1, 1, 4, 8, 8, 1],[0, 0, 8, 4, 0, 0, 0, 8, 4, 0],[3, 1, 2, 3, 5, 3, 1, 2, 3, 5],[9, 6, 5, 8, 0, 9, 6, 5, 8, 0],[3, 6, 1, 7, 2, 3, 6, 1, 7, 2]]) import matplotlib.pyplot as plt lena = plt.imread('lena.jpg') lenas = np.concatenate((lena, lena)) plt.imshow(lenas) plt.show()

np.hstack與np.vstack

水平級聯(lián)與垂直級聯(lián), 處理自己, 進行維度的變更

# vertical 垂直 n10 = np.random.randint(150, size=10) n10 array([138, 122, 78, 145, 95, 114, 98, 49, 137, 147]) n11 = np.vstack(n10) n11 array([[138],[122],[ 78],[145],[ 95],[114],[ 98],[ 49],[137],[147]]) n12 = np.array([[2, 3, 4, 5, 6]]) # 可以對二維及多位數(shù)組進行降維 np.hstack(n12) array([2, 3, 4, 5, 6]) n13 = np.array([[2, 3, 4, 5, 6], [4, 5, 6, 7, 8]]) np.hstack(n13) array([2, 3, 4, 5, 6, 4, 5, 6, 7, 8]) np.hstack(np.hstack(lena)) array([225, 138, 128, ..., 188, 74, 84], dtype=uint8)

切分

與級聯(lián)類似, 三個函數(shù)完成切分:
- np.split
- np.vsplit
- np.hsplit

# np.split(ary, indices_or_sections, axis=0) n14 = np.random.randint(0, 150, size=(5, 7)) n14 array([[ 0, 107, 40, 62, 108, 120, 130],[ 40, 79, 34, 48, 110, 48, 24],[ 52, 121, 69, 18, 88, 73, 64],[147, 41, 118, 138, 128, 69, 76],[ 67, 30, 77, 87, 10, 18, 69]]) # 第一行切第三行切 np.split(n14, (1, 3)) [array([[ 0, 107, 40, 62, 108, 120, 130]]),array([[ 40, 79, 34, 48, 110, 48, 24],[ 52, 121, 69, 18, 88, 73, 64]]),array([[147, 41, 118, 138, 128, 69, 76],[ 67, 30, 77, 87, 10, 18, 69]])] lena3 = np.split(lena, (200, 350))[1] plt.imshow(lena3) plt.show()

# axis=0 默認, 切分行 # axis=1, 切分列 np.split(n14, (1, 3), axis=1) [array([[ 0],[ 40],[ 52],[147],[ 67]]), array([[107, 40],[ 79, 34],[121, 69],[ 41, 118],[ 30, 77]]), array([[ 62, 108, 120, 130],[ 48, 110, 48, 24],[ 18, 88, 73, 64],[138, 128, 69, 76],[ 87, 10, 18, 69]])] # 豎直方向切分的是行 np.vsplit(n14, (1, 3)) [array([[ 0, 107, 40, 62, 108, 120, 130]]),array([[ 40, 79, 34, 48, 110, 48, 24],[ 52, 121, 69, 18, 88, 73, 64]]),array([[147, 41, 118, 138, 128, 69, 76],[ 67, 30, 77, 87, 10, 18, 69]])] # 水平方向切分的是列 np.hsplit(n14, (2, 3)) [array([[ 0, 107],[ 40, 79],[ 52, 121],[147, 41],[ 67, 30]]), array([[ 40],[ 34],[ 69],[118],[ 77]]), array([[ 62, 108, 120, 130],[ 48, 110, 48, 24],[ 18, 88, 73, 64],[138, 128, 69, 76],[ 87, 10, 18, 69]])]

副本

所有賦值運算不會為ndarray的任何元素創(chuàng)建副本, 對賦值后的對象的操作也對原來的對象生效.

l = [1, 2, 3, 4] n = np.array(l) n array([1, 2, 3, 4]) # 當數(shù)據(jù)是ndarray時, 如果用=賦值, 內(nèi)存沒有改變 n2 = n n2[2] = 90 n2 array([ 1, 2, 90, 4]) n array([ 1, 2, 90, 4]) # copy()函數(shù)創(chuàng)建副本 n3 = n.copy() n3[0] = 80 n3 array([80, 2, 90, 4]) n array([ 1, 2, 90, 4])

ndarray的聚合操作

求和 np.sum

n15.mean() 87.0625 n15.mean(axis=0) array([[ 82.75, 85.25, 58.5 , 116. ],[ 69. , 96.5 , 56. , 115.25],[ 95.5 , 92.5 , 107.75, 108.5 ],[ 47.25, 95.5 , 47. , 119.75]]) import numpy as np import matplotlib.pyplot as plt n16 = np.random.randint(0, 150, size=(4, 4, 4)) n16 array([[[114, 88, 143, 100],[ 91, 7, 84, 49],[114, 54, 14, 20],[ 83, 12, 135, 1]],[[ 36, 95, 80, 96],[ 97, 98, 39, 146],[ 2, 127, 53, 105],[ 71, 15, 11, 97]],[[ 3, 64, 27, 2],[109, 28, 81, 123],[ 64, 95, 112, 66],[ 42, 131, 79, 123]],[[ 54, 31, 10, 133],[138, 23, 145, 122],[ 66, 29, 79, 97],[119, 139, 12, 100]]]) np.mean(n16, axis=0) array([[ 51.75, 69.5 , 65. , 82.75],[108.75, 39. , 87.25, 110. ],[ 61.5 , 76.25, 64.5 , 72. ],[ 78.75, 74.25, 59.25, 80.25]]) np.sum(n16, axis=0) array([[207, 278, 260, 331],[435, 156, 349, 440],[246, 305, 258, 288],[315, 297, 237, 321]])

最大值和最小值 np.max/np.min

display(lena.max(), lena.min()) 2550 display(lena.max(axis=0), lena.min()) array([[240, 155, 139],[239, 152, 138],[240, 154, 135],...,[244, 211, 177],[246, 211, 178],[245, 209, 179]], dtype=uint8)0 n15 = np.random.randint(0, 150, size=(4, 4, 4)) n15 array([[[ 18, 91, 115, 148],[141, 145, 58, 148],[ 80, 97, 70, 82],[ 48, 47, 85, 108]],[[ 91, 87, 13, 93],[ 18, 50, 7, 145],[124, 74, 124, 105],[ 7, 124, 29, 130]],[[146, 77, 76, 98],[ 34, 126, 34, 96],[127, 70, 148, 131],[ 25, 148, 2, 141]],[[ 76, 86, 30, 125],[ 83, 65, 125, 72],[ 51, 129, 89, 116],[109, 63, 72, 100]]]) n15.max() 148 n15.max(axis=0) array([[146, 91, 115, 148],[141, 145, 125, 148],[127, 129, 148, 131],[109, 148, 85, 141]]) n15.max(axis=2) array([[148, 148, 97, 108],[ 93, 145, 124, 130],[146, 126, 148, 148],[125, 125, 129, 109]])

其他聚合操作

np.std: 標準方差
np.power: 冪運算
np.argmin: 最小值的索引
np.argmax: 最大值的索引
np.argwhere: 滿足條件的元素的索引
np.sum和np.nansum的區(qū)別: nan not a number

n20 = np.random.randint(0, 100, size=10) n20 array([35, 84, 96, 92, 32, 96, 76, 72, 82, 59]) np.argmin(n20) 4 np.argmax(n20) 2 np.argwhere(n20 > 70) array([[1],[2],[3],[5],[6],[7],[8]], dtype=int64) index = np.argwhere(n20 > 70) n20[index] array([[84],[96],[92],[96],[76],[72],[82]]) n20[np.array([[0], [1]])] array([[35],[84]])

操作文件

使用pandas打開文件.csv 獲取文件中的數(shù)據(jù)

import pandas as pd df = pd.read_csv('../data/height.csv') df ordernameheight012345

1	Jay	175
2	JJ	175
3	four	168
4	Neng	170
5	Xie	165
6	Feet	170

df.values array([[1, 'Jay', 175],[2, 'JJ', 175],[3, 'four', 168],[4, 'Neng', 170],[5, 'Xie', 165],[6, 'Feet', 170]], dtype=object)

ndarray的矩陣操作

基本矩陣操作

算術運算符

n21 = np.random.randint(0, 10, size=(4, 5)) n21 array([[4, 9, 1, 9, 9],[7, 7, 3, 7, 3],[6, 4, 2, 5, 4],[0, 0, 1, 4, 8]]) n21 + 10 array([[14, 19, 11, 19, 19],[17, 17, 13, 17, 13],[16, 14, 12, 15, 14],[10, 10, 11, 14, 18]]) n22 = n21 / 2 n22 array([[2. , 4.5, 0.5, 4.5, 4.5],[3.5, 3.5, 1.5, 3.5, 1.5],[3. , 2. , 1. , 2.5, 2. ],[0. , 0. , 0.5, 2. , 4. ]]) np.add(n21, n21) array([[ 8, 18, 2, 18, 18],[14, 14, 6, 14, 6],[12, 8, 4, 10, 8],[ 0, 0, 2, 8, 16]])

矩陣積 np.dot()

n23 = np.random.randint(0, 10, size=(2, 3)) n24 = np.random.randint(0, 10, size=(3, 2)) display(n23, n24) array([[2, 2, 4],[8, 7, 8]])array([[3, 0],[2, 1],[3, 0]]) np.dot(n23, n24) array([[22, 2],[62, 7]])

廣播機制

ndarray廣播機制的兩條規(guī)則:
1. 為缺失的維度補1
2. 嘉定缺失元素用已有值填充

m = np.ones((2, 3)) a = np.arange(3) display(m, a) array([[1., 1., 1.],[1., 1., 1.]])array([0, 1, 2]) # numpy的廣播機制, 維度不對應, 自動補全 m + a array([[1., 2., 3.],[1., 2., 3.]]) b = np.arange(3).reshape((3, 1)) b1 = np.arange(3) display(b, b1) b + b1 array([[0],[1],[2]])array([0, 1, 2])array([[0, 1, 2],[1, 2, 3],[2, 3, 4]]) b2 = np.ones((4, 1)) b3 = np.arange(4) display(b2, b3) array([[1.],[1.],[1.],[1.]])array([0, 1, 2, 3]) b2 + b3 array([[1., 2., 3., 4.],[1., 2., 3., 4.],[1., 2., 3., 4.],[1., 2., 3., 4.]])

ndarray的排序

n31 = np.array([2, 5, 1, 7, 4]) def sortn(nd):'''冒泡排序'''for i in range(nd.size):for j in range(i, nd.size):if nd[i] > nd[j]:nd[i], nd[j] = nd[j], nd[i]return nd sortn(n31) array([1, 2, 4, 5, 7]) # 降低運算的空間復雜度和時間復雜度 def sortnd(nd):for i in range(nd.size): # 切片, 索引不對應min_index = np.argmin(nd[i:]) + i # print(min_index) # print(i,nd[i],nd[min_index])nd[i], nd[min_index] = nd[min_index], nd[i]return nd sortnd(n31) array([1, 2, 4, 5, 7])

快速排序

np.sort()與ndarray.sort()都可以, 但是有區(qū)別:
- np.sort()不改變輸入
- ndarray.sort()本地處理, 不占用空間, 但改變輸入

# a.sort(axis=-1, kind='quicksort', order=None) n32 = np.random.randint(0, 150, size=10) n32 array([131, 31, 101, 57, 96, 50, 142, 133, 83, 141]) # 使用ndarray.sort(), 原來的數(shù)據(jù)進行了改變, 不占內(nèi)存 n32.sort() n32 array([ 0, 24, 36, 43, 85, 121, 121, 135, 138, 141]) n33 = np.sort(n32) display(n32, n33) array([131, 31, 101, 57, 96, 50, 142, 133, 83, 141])array([ 31, 50, 57, 83, 96, 101, 131, 133, 141, 142])

部分排序

np.partition(a, k)

當k為正時, 得到最小的k個數(shù)
當k為負時, 得到最大的k個數(shù)

nd = np.random.randint(0, 150, size=20) nd array([145, 134, 88, 140, 10, 59, 132, 134, 31, 74, 91, 79, 18,44, 21, 140, 34, 89, 63, 26]) np.partition(nd,-5) array([ 18, 26, 63, 34, 10, 59, 21, 44, 31, 74, 79, 91, 88,89, 132, 134, 134, 140, 140, 145]) np.partition(nd,5) array([ 10, 18, 21, 26, 31, 34, 44, 59, 63, 74, 91, 79, 134,134, 132, 140, 140, 89, 88, 145])

轉載于:https://www.cnblogs.com/KX-Lau/p/10331747.html

總結

以上是生活随笔為你收集整理的numpy---(精简)的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。

NumPy