NumpyExampleList

ThispagecontainsalargedatabaseofexamplesdemonstratingmostoftheNumpy
functionality.NumpyExampleListWithDochastheseexamplesinterleavedwiththe
builtindocumentation,butisnotasregularlyupdatedasthispage.Theexampleshere
canbeeasilyaccessedfromPythonusingtheNumpyExampleFetcher.
Thisexamplelistisincrediblyuseful,andwewouldliketogetallthegoodexamples
andcommentsintegratedintheofficialnumpydocumentationsothattheyarealso
shippedwithnumpy.Youcanhelp.Theofficialnumpydocumentationcanbeedited
onhttp://docs.scipy.org.
TabladeContenidos
1. ...
2. []
3. abs()
4. absolute()
5. accumulate()
6. add()
7. all()
8. allclose()
9. alltrue()
10. angle()
11. any()
12. append()
13. apply_along_axis()
14. apply_over_axes()
15. arange()
16. arccos()
17. arccosh()
18. arcsin()
19. arcsinh()
20. arctan()
21. arctan2()
22. arctanh()
23. argmax()
24. argmin()
25. argsort()
26. array()
27. arrayrange()
28. array_split()
29. asarray()
30. asanyarray()

31. asmatrix()
32. astype()
33. atleast_1d()
34. atleast_2d()
35. atleast_3d()
36. average()
37. beta()
38. binary_repr()
39. bincount()
40. binomial()
41. bitwise_and()
42. bitwise_or()
43. bitwise_xor()
44. bmat()
45. broadcast()
46. bytes()
47. c_[]
48. cast[]()
49. ceil()
50. choose()
51. clip()
52. column_stack()
53. compress()
54. concatenate()
55. conj()
56. conjugate()
57. copy()
58. corrcoef()
59. cos()
60. cov()
61. cross()
62. cumprod()
63. cumsum()
64. delete()
65. det()
66. diag()
67. diagflat()
68. diagonal()
69. diff()
70. digitize()
71. dot()
72. dsplit()

73. dstack()
74. dtype()
75. empty()
76. empty_like()
77. expand_dims()
78. eye()
79. fft()
80. fftfreq()
81. fftshift()
82. fill()
83. finfo()
84. fix()
85. flat
86. flatten()
87. fliplr()
88. flipud()
89. floor()
90. fromarrays()
91. frombuffer()
92. fromfile()
93. fromfunction()
94. fromiter()
95. generic
96. gumbel()
97. histogram()
98. hsplit()
99. hstack()
100. hypot()
101. identity()
102. ifft()
103. imag
104. index_exp[]
105. indices()
106. inf
107. inner()
108. insert()
109. inv()
110. iscomplex()
111. iscomplexobj()
112. item()
113. ix_()
114. lexsort()

115. linspace()
116. loadtxt()
117. logical_and()
118. logical_not()
119. logical_or()
120. logical_xor()
121. logspace()
122. lstsq()
123. mat()
124. matrix()
125. max()
126. maximum()
127. mean()
128. median()
129. mgrid[]
130. min()
131. minimum()
132. multiply()
133. nan
134. ndenumerate()
135. ndim
136. ndindex()
137. newaxis
138. nonzero()
139. ogrid()
140. ones()
141. ones_like()
142. outer()
143. permutation()
144. piecewise()
145. pinv()
146. poisson()
147. poly1d()
148. polyfit()
149. prod()
150. ptp()
151. put()
152. putmask()
153. r_[]
154. rand()
155. randint()
156. randn()

157. random_integers()
158. random_sample()
159. ranf()
160. ravel()
161. real
162. recarray()
163. reduce()
164. repeat()
165. reshape()
166. resize()
167. rollaxis()
168. round()
169. rot90()
170. s_[]
171. sample()
172. savetxt()
173. searchsorted()
174. seed()
175. select()
176. set_printoptions()
177. shape
178. shuffle()
179. slice()
180. solve()
181. sometrue()
182. sort()
183. split()
184. squeeze()
185. std()
186. standard_normal()
187. sum()
188. svd()
189. swapaxes()
190. T
191. take()
192. tensordot()
193. tile()
194. tofile()
195. tolist()
196. trace()
197. transpose()
198. tri()

199. tril()
200. trim_zeros()
201. triu()
202. typeDict()
203. uniform()
204. unique()
205. unique1d()
206. vander()
207. var()
208. vdot()
209. vectorize()
210. view()
211. vonmises()
212. vsplit()
213. vstack()
214. weibull()
215. where()
216. zeros()
217. zeros_like()

...
>>>fromnumpyimport*
>>>a=arange(12)
>>>a=a.reshape(3,2,2)
>>>printa
[[[01]
[23]]
[[45]
[67]]
[[89]
[1011]]]
>>>a[...,0]#sameasa[:,:,0]
array([[0,2],
[4,6],
[8,10]])
>>>a[1:,...]#sameasa[1:,:,:]orjusta[1:]
array([[[4,5],
[6,7]],
[[8,9],
[10,11]]])

Seealso:[],newaxis

[]

>>>fromnumpyimport*
>>>a=array([[0,1,2,3,4],
...[10,11,12,13,14],
...[20,21,22,23,24],
...[30,31,32,33,34]])
>>>
>>>a[0,0]#indicesstartbyzero
0
>>>a[1]#lastrow
array([30,31,32,33,34])
>>>a[1:3,1:4]#subarray
array([[11,12,13],
[21,22,23]])
>>>
>>>i=array([0,1,2,1])#arrayofindicesforthefirstaxis
>>>j=array([1,2,3,4])#arrayofindicesforthesecondaxis
>>>a[i,j]
array([1,12,23,14])
>>>
>>>a[a<13]#booleanindexing
array([0,1,2,3,4,10,11,12])
>>>
>>>b1=array([True,False,True,False])#booleanrowselector
>>>a[b1,:]
array([[0,1,2,3,4],
[20,21,22,23,24]])
>>>
>>>b2=array([False,True,True,False,True])#booleancolumn
selector
>>>a[:,b2]
array([[1,2,4],
[11,12,14],
[21,22,24],
[31,32,34]])

Seealso:...,newaxis,ix_,indices,nonzero,where,slice

abs()
>>>fromnumpyimport*
>>>abs(1)
1
>>>abs(array([1.2,1.2]))
array([1.2,1.2])
>>>abs(1.2+1j)
1.5620499351813308

Seealso:absolute,angle

absolute()
Synonymforabs()
Seeabs

accumulate()
>>>fromnumpyimport*
>>>add.accumulate(array([1.,2.,3.,4.]))#likereduce()butalso
givesintermediateresults
array([1.,3.,6.,10.])
>>>array([1.,1.+2.,(1.+2.)+3.,((1.+2.)+3.)+4.])#thisiswhat
itcomputed
array([1.,3.,6.,10.])
>>>multiply.accumulate(array([1.,2.,3.,4.]))#worksalsowith
otheroperands
array([1.,2.,6.,24.])
>>>array([1.,1.*2.,(1.*2.)*3.,((1.*2.)*3.)*4.])#thisiswhat
itcomputed
array([1.,2.,6.,24.])
>>>add.accumulate(array([[1,2,3],[4,5,6]]),axis=0)#
accumulateeverycolumnseparately
array([[1,2,3],
[5,7,9]])
>>>add.accumulate(array([[1,2,3],[4,5,6]]),axis=1)#
accumulateeveryrowseparately
array([[1,3,6],
[4,9,15]])

Seealso:reduce,cumprod,cumsum

add()
>>>fromnumpyimport*
>>>add(array([1.2,1.2]),array([1,3]))
array([0.2,4.2])
>>>array([1.2,1.2])+array([1,3])
array([0.2,4.2])

all()
>>>fromnumpyimport*
>>>a=array([True,False,True])
>>>a.all()#ifallelementsofaareTrue:returnTrue
otherwiseFalse
False
>>>all(a)#thisformalsoexists
False

>>>a=array([1,2,3])
>>>all(a>0)#equivalentto(a>0).all()
True

Seealso:any,alltrue,sometrue

allclose()
>>>allclose(array([1e10,1e7]),array([1.00001e10,1e8]))
False
>>>allclose(array([1e10,1e8]),array([1.00001e10,1e9]))
True
>>>allclose(array([1e10,1e8]),array([1.0001e10,1e9]))
False

alltrue()
>>>fromnumpyimport*
>>>b=array([True,False,True,True])
>>>alltrue(b)
False
>>>a=array([1,5,2,7])
>>>alltrue(a>=5)
False

Seealso:sometrue,all,any

angle()
>>>fromnumpyimport*
>>>angle(1+1j)#inradians
0.78539816339744828
>>>angle(1+1j,deg=True)#indegrees
45.0

Seealso:real,imag,hypot

any()
>>>fromnumpyimport*
>>>a=array([True,False,True])
>>>a.any()#givesTrueifatleast1elementofaisTrue,
otherwiseFalse
True
>>>any(a)#thisformalsoexists
True
>>>a=array([1,2,3])
>>>(a>=1).any()#equivalenttoany(a>=1)

True

Seealso:all,alltrue,sometrue

append()
>>>fromnumpyimport*
>>>a=array([10,20,30,40])
>>>append(a,50)
array([10,20,30,40,50])
>>>append(a,[50,60])
array([10,20,30,40,50,60])
>>>a=array([[10,20,30],[40,50,60],[70,80,90]])
>>>append(a,[[15,15,15]],axis=0)
array([[10,20,30],
[40,50,60],
[70,80,90],
[15,15,15]])
>>>append(a,[[15],[15],[15]],axis=1)
array([[10,20,30,15],
[40,50,60,15],
[70,80,90,15]])

Seealso:insert,delete,concatenate

apply_along_axis()
>>>fromnumpyimport*
>>>defmyfunc(a):#functionworksona1darrays,takesthe
averageofthe1stanlastelement
...return(a[0]+a[1])/2
...
>>>b=array([[1,2,3],[4,5,6],[7,8,9]])
>>>apply_along_axis(myfunc,0,b)#applymyfunctoeachcolumn
(axis=0)ofb
array([4,5,6])
>>>apply_along_axis(myfunc,1,b)#applymyfunctoeachrow
(axis=1)ofb
array([2,5,8])

Seealso:apply_over_axes,vectorize

apply_over_axes()
>>>fromnumpyimport*
>>>a=arange(24).reshape(2,3,4)#ahas3axes:0,1and2
>>>a
array([[[0,1,2,3],
[4,5,6,7],

[8,9,10,11]],
[[12,13,14,15],
[16,17,18,19],
[20,21,22,23]]])
>>>apply_over_axes(sum,a,[0,2])#sumoverallaxesexcept
axis=1,resulthassameshapeasoriginal
array([[[60],
[92],
[124]]])

Seealso:apply_along_axis,vectorize

arange()
>>>fromnumpyimport*
>>>arange(3)
array([0,1,2])
>>>arange(3.0)
array([0.,1.,2.])
>>>arange(3,dtype=float)
array([0.,1.,2.])
>>>arange(3,10)#start,stop
array([3,4,5,6,7,8,9])
>>>arange(3,10,2)#start,stop,step
array([3,5,7,9])

Seealso:r_,linspace,logspace,mgrid,ogrid

arccos()
>>>fromnumpyimport*
>>>arccos(array([0,1]))
array([1.57079633,0.])

Seealso:arcsin,arccosh,arctan,arctan2

arccosh()
>>>fromnumpyimport*
>>>arccosh(array([e,10.0]))
array([1.65745445,2.99322285])

Seealso:arccos,arcsinh,arctanh

arcsin()
>>>fromnumpyimport*
>>>arcsin(array([0,1]))

array([0.,1.57079633])

Seealso:arccos,arctan,arcsinh

arcsinh()
>>>fromnumpyimport*
>>>arcsinh(array([e,10.0]))
array([1.72538256,2.99822295])

Seealso:arccosh,arcsin,arctanh

arctan()
>>>fromnumpyimport*
>>>arctan(array([0,1]))
array([0.,0.78539816])

Seealso:arccos,arcsin,arctanh

arctan2()
>>>fromnumpyimport*
>>>arctan2(array([0,1]),array([1,0]))
array([0.,1.57079633])

Seealso:arcsin,arccos,arctan,arctanh

arctanh()
>>>fromnumpyimport*
>>>arctanh(array([0,0.5]))
array([0.,0.54930614])

Seealso:arcsinh,arccosh,arctan,arctan2

argmax()
>>>fromnumpyimport*
>>>a=array([10,20,30])
>>>maxindex=a.argmax()
>>>a[maxindex]
30
>>>a=array([[10,50,30],[60,20,40]])
>>>maxindex=a.argmax()
>>>maxindex
3

>>>a.ravel()[maxindex]
60
>>>a.argmax(axis=0)#foreachcolumn:therowindexofthe
maximumvalue
array([1,0,1])
>>>a.argmax(axis=1)#foreachrow:thecolumnindexofthe
maximumvalue
array([1,0])
>>>argmax(a)#alsoexists,slower,defaultisaxis=1
array([1,0])

Seealso:argmin,nan,min,max,maximum,minimum

argmin()
>>>fromnumpyimport*
>>>a=array([10,20,30])
>>>minindex=a.argmin()
>>>a[minindex]
10
>>>a=array([[10,50,30],[60,20,40]])
>>>minindex=a.argmin()
>>>minindex
0
>>>a.ravel()[minindex]
10
>>>a.argmin(axis=0)#foreachcolumn:therowindexofthe
minimumvalue
array([0,1,0])
>>>a.argmin(axis=1)#foreachrow:thecolumnindexofthe
minimumvalue
array([0,1])
>>>argmin(a)#alsoexists,slower,defaultisaxis=1
array([0,1])

Seealso:argmax,nan,min,max,maximum,minimum

argsort()
argsort(axis=1,kind="quicksort")
>>>fromnumpyimport*
>>>a=array([2,0,8,4,1])
>>>ind=a.argsort()#indicesofsortedarrayusingquicksort
(default)
>>>ind
array([1,4,0,3,2])
>>>a[ind]#sameeffectasa.sort()
array([0,1,2,4,8])
>>>ind=a.argsort(kind='merge')#algorithmoptionsare

'quicksort','mergesort'and'heapsort'
>>>a=array([[8,4,1],[2,0,9]])
>>>ind=a.argsort(axis=0)#sortsoncolumns.NOTthesameas
a.sort(axis=1)
>>>ind
array([[1,1,0],
[0,0,1]])
>>>a[ind,[[0,1,2],[0,1,2]]]#2Darraysneedfancyindexingif
youwanttosortthem.
array([[2,0,1],
[8,4,9]])
>>>ind=a.argsort(axis=1)#sortalongrows.Canuse
a.argsort(axis=1)forlastaxis.
>>>ind
array([[2,1,0],
[1,0,2]])
>>>a=ones(17)
>>>a.argsort()#quicksortdoesn'tpreserveoriginalorder.
array([0,14,13,12,11,10,9,15,8,6,5,4,3,2,1,7,16])
>>>a.argsort(kind="mergesort")#mergesortpreservesorderwhen
possible.Itisastablesort.
array([0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16])
>>>ind=argsort(a)#thereisafunctionalform

Seealso:lexsort,sort

array()
>>>fromnumpyimport*
>>>array([1,2,3])#conversionfromalisttoanarray
array([1,2,3])
>>>array([1,2,3],dtype=complex)#outputtypeisspecified
array([1.+0.j,2.+0.j,3.+0.j])
>>>array(1,copy=0,subok=1,ndmin=1)#basicallyequivalentto
atleast_1d
array([1])
>>>array(1,copy=0,subok=1,ndmin=2)#basicallyequivalentto
atleast_2d
array([[1]])
>>>array(1,subok=1,ndmin=2)#likeatleast_2dbutalwaysmakes
acopy
array([[1]])
>>>mydescriptor={'names':('gender','age','weight'),'formats':
('S1','f4','f4')}#onewayofspecifyingthedatatype
>>>a=array([('M',64.0,75.0),('F',25.0,60.0)],
dtype=mydescriptor)#recarray
>>>printa
[('M',64.0,75.0)('F',25.0,60.0)]
>>>a['weight']
array([75.,60.],dtype=float32)

>>>a.dtype.names#Accesstotheorderedfieldnames
('gender','age','weight')
>>>mydescriptor=[('age',int16),('Nchildren',int8),
('weight',float32)]#anotherwayofspecifyingthedatatype
>>>a=array([(64,2,75.0),(25,0,60.0)],dtype=mydescriptor)
>>>a['Nchildren']
array([2,0],dtype=int8)
>>>mydescriptor=dtype([('x','f4'),('y','f4'),#nested
recarray
...('nested',[('i','i2'),('j','i2')])])
>>>array([(1.0,2.0,(1,2))],dtype=mydescriptor)#inputonerow
array([(1.0,2.0,(1,2))],
dtype=[('x','<f4'),('y','<f4'),('nested',[('i','<i2'),
('j','<i2')])])
>>>array([(1.0,2.0,(1,2)),(2.1,3.2,(3,2))],
dtype=mydescriptor)#inputtworows
array([(1.0,2.0,(1,2)),(2.0999999046325684,
3.2000000476837158,(3,2))],
dtype=[('x','<f4'),('y','<f4'),('nested',[('i','<i2'),
('j','<i2')])])
>>>a=array([(1.0,2.0,(1,2)),(2.1,3.2,(3,2))],
dtype=mydescriptor)#gettingsomecolumns
>>>a['x']#aplaincolumn
array([1.,2.0999999],dtype=float32)
>>>a['nested']#anestedcolumn
array([(1,2),(3,2)],
dtype=[('i','<i2'),('j','<i2')])
>>>a['nested']['i']#aplaincolumninsideanestedcolumn
>>>mydescriptor=dtype([('x','f4'),('y','f4'),#nested
recarray
...('nested',[('i','i2'),('j','i2')])])
>>>array([(1.0,2.0,(1,2))],dtype=mydescriptor)#inputonerow
array([(1.0,2.0,(1,2))],
dtype=[('x','<f4'),('y','<f4'),('nested',[('i','<i2'),
('j','<i2')])])
>>>array([(1.0,2.0,(1,2)),(2.1,3.2,(3,2))],
dtype=mydescriptor)#inputtworows
array([(1.0,2.0,(1,2)),(2.0999999046325684,
3.2000000476837158,(3,2))],
dtype=[('x','<f4'),('y','<f4'),('nested',[('i','<i2'),
('j','<i2')])])
>>>a=array([(1.0,2.0,(1,2)),(2.1,3.2,(3,2))],
dtype=mydescriptor)#gettingsomecolumns
>>>a['x']#aplaincolumn
array([1.,2.0999999],dtype=float32)
>>>a['nested']#anestedcolumn
array([(1,2),(3,2)],
dtype=[('i','<i2'),('j','<i2')])
>>>a['nested']['i']#aplaincolumninsideanestedcolumn
array([1,3],dtype=int16)

Seealso:dtype,mat,asarray

arrayrange()
Synonymforarange()
Seearange

array_split()
>>>fromnumpyimport*
>>>a=array([[1,2,3,4],[5,6,7,8]])
>>>array_split(a,2,axis=0)#splitain2parts.rowwise
[array([[1,2,3,4]]),array([[5,6,7,8]])]
>>>array_split(a,4,axis=1)#splitain4parts,columnwise
[array([[1],
[5]]),array([[2],
[6]]),array([[3],
[7]]),array([[4],
[8]])]
>>>array_split(a,3,axis=1)#impossibletosplitin3equal
parts>firstpart(s)arebigger
[array([[1,2],
[5,6]]),array([[3],
[7]]),array([[4],
[8]])]
>>>array_split(a,[2,3],axis=1)#makeasplitbeforethe2ndand
the3rdcolumn
[array([[1,2],
[5,6]]),array([[3],
[7]]),array([[4],
[8]])]

Seealso:dsplit,hsplit,vsplit,split,concatenate

asarray()
>>>fromnumpyimport*
>>>m=matrix('1258')
>>>m
matrix([[1,2],
[5,8]])
>>>a=asarray(m)#aisarraytypewithsamecontentsasm
dataisnotcopied
>>>a
array([[1,2],
[5,8]])
>>>m[0,0]=99
>>>m
matrix([[99,2],
[5,8]])

>>>a#nocopywasmade,somodifyingmmodifiesa,andvice
versa
array([[99,2],
[5,8]])

Seealso:asmatrix,array,matrix,mat

asanyarray()
>>>fromnumpyimport*
>>>a=array([[1,2],[5,8]])
>>>a
array([[1,2],
[5,8]])
>>>m=matrix('1258')
>>>m
matrix([[1,2],
[5,8]])
>>>asanyarray(a)#thearrayaisreturnedunmodified
array([[1,2],
[5,8]])
>>>asanyarray(m)#thematrixmisreturnedunmodified
matrix([[1,2],
[5,8]])
>>>asanyarray([1,2,3])#anewarrayisconstructedfromthelist
array([1,2,3])

Seealso:asmatrix,asarray,array,mat

asmatrix()
>>>fromnumpyimport*
>>>a=array([[1,2],[5,8]])
>>>a
array([[1,2],
[5,8]])
>>>m=asmatrix(a)#mismatrixtypewithsamecontentsasa
dataisnotcopied
>>>m
matrix([[1,2],
[5,8]])
>>>a[0,0]=99
>>>a
array([[99,2],
[5,8]])
>>>m#nocopywasmadesomodifyingamodifiesm,andviceversa
matrix([[99,2],
[5,8]])

Seealso:asarray,array,matrix,mat

astype()
>>>fromnumpyimport*
>>>x=array([1,2,3])
>>>y=x.astype(float64)#convertfromint32tofloat64
>>>type(y[0])
<type'numpy.float64'>
>>>x.astype(None)#Noneimpliesconvertingtothedefault
(float64)
array([1.,2.,3.])

Seealso:cast,dtype,ceil,floor,round_,fix

atleast_1d()
>>>fromnumpyimport*
>>>a=1#0darray
>>>b=array([2,3])#1darray
>>>c=array([[4,5],[6,7]])#2darray
>>>d=arange(8).reshape(2,2,2)#3darray
>>>d
array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])
>>>atleast_1d(a,b,c,d)#alloutputarrayshavedim>=1
[array([1]),array([2,3]),array([[4,5],
[6,7]]),array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])]

Seealso:atleast_2d,atleast_3d,newaxis,expand_dims

atleast_2d()
>>>fromnumpyimport*
>>>a=1#0darray
>>>b=array([2,3])#1darray
>>>c=array([[4,5],[6,7]])#2darray
>>>d=arange(8).reshape(2,2,2)#3darray
>>>d
array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])
>>>atleast_2d(a,b,c,d)#alloutputarrayshavedim>=2

[array([[1]]),array([[2,3]]),array([[4,5],
[6,7]]),array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])]

Seealso:atleast_1d,atleast_3d,newaxis,expand_dims

atleast_3d()
>>>fromnumpyimport*
>>>a=1#0darray
>>>b=array([2,3])#1darray
>>>c=array([[4,5],[6,7]])#2darray
>>>d=arange(8).reshape(2,2,2)#3darray
>>>d
array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])
>>>atleast_3d(a,b,c,d)#alloutputarrayshavedim>=3
[array([[[1]]]),array([[[2],
[3]]]),array([[[4],
[5]],
[[6],
[7]]]),array([[[0,1],
[2,3]],
[[4,5],
[6,7]]])]

Seealso:atleast_1d,atleast_2d,newaxis,expand_dims

average()
>>>fromnumpyimport*
>>>a=array([1,2,3,4,5])
>>>w=array([0.1,0.2,0.5,0.2,0.2])#weights,not
necessarilynormalized
>>>average(a)#plainmeanvalue
3.0
>>>average(a,weights=w)#weightedaverage
3.1666666666666665
>>>average(a,weights=w,returned=True)#output=weighted
average,sumofweights
(3.1666666666666665,1.2)

Seealso:mean,median

beta()

>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>beta(a=1,b=10,size=(2,2))#Betadistributionalpha=1,beta=10
array([[0.02571091,0.04973536],
[0.04887027,0.02382052]])

Seealso:seed

binary_repr()
>>>fromnumpyimport*
>>>a=25
>>>binary_repr(a)#binaryrepresentationof25
'11001'
>>>b=float_(pi)#numpyfloathasextrafunctionality...
>>>b.nbytes#...likethenumberofbytesittakes
8
>>>binary_repr(b.view('u8'))#viewfloatnumberasan8byte
integer,thengetbinarybitstring
'1010100010001000010110100011000'

bincount()
>>>fromnumpyimport*
>>>a=array([1,1,1,1,2,2,4,4,5,6,6,6])#doesn'tneedtobe
sorted
>>>bincount(a)#0occurs0times,1occurs4times,2occurs
twice,3occurs0times,...
array([0,4,2,0,2,1,3])
>>>a=array([5,4,4,2,2])
>>>w=array([0.1,0.2,0.1,0.3,0.5])
>>>bincount(a)#0&1don'toccur,2occurstwice,3doesn't
occur,4occurstwice,5once
array([0,0,2,0,2,1])
>>>bincount(a,weights=w)
array([0.,0.,0.8,0.,0.3,0.1])
>>>#0occurs0times>result[0]=0
>>>#1occurs0times>result[1]=0
>>>#2occursatindices3&4>result[2]=w[3]+w[4]
>>>#3occurs0times>result[3]=0
>>>#4occursatindices1&2>result[4]=w[1]+w[2]
>>>#5occursatindex0>result[5]=w[0]

Seealso:histogram,digitize

binomial()
>>>fromnumpyimport*

>>>fromnumpy.randomimport*
>>>binomial(n=100,p=0.5,size=(2,3))#binomialdistributionn
trials,p=successprobability
array([[38,50,53],
[56,48,54]])
>>>frompylabimport*#histogramplotexample
>>>hist(binomial(100,0.5,(1000)),20)

Seealso:random_sample,uniform,standard_normal,seed

bitwise_and()
>>>fromnumpyimport*
>>>bitwise_and(array([2,5,255]),array([4,4,4]))
array([0,4,4])
>>>bitwise_and(array([2,5,255,2147483647L],dtype=int32),
array([4,4,4,2147483647L],dtype=int32))
array([0,4,4,2147483647])

Seealso:bitwise_or,bitwise_xor,logical_and

bitwise_or()
>>>fromnumpyimport*
>>>bitwise_or(array([2,5,255]),array([4,4,4]))
array([6,5,255])
>>>bitwise_or(array([2,5,255,2147483647L],dtype=int32),
array([4,4,4,2147483647L],dtype=int32))
array([6,5,255,2147483647])

Seealso:bitwise_and,bitwise_xor,logical_or

bitwise_xor()
>>>fromnumpyimport*
>>>bitwise_xor(array([2,5,255]),array([4,4,4]))
array([6,1,251])
>>>bitwise_xor(array([2,5,255,2147483647L],dtype=int32),
array([4,4,4,2147483647L],dtype=int32))
array([6,1,251,0])

Seealso:bitwise_and,bitwise_or,logical_xor

bmat()
>>>fromnumpyimport*
>>>a=mat('1234')
>>>b=mat('5678')

>>>bmat('abba')#allelementsmustbeexistingsymbols
matrix([[1,2,5,6],
[3,4,7,8],
[5,6,1,2],
[7,8,3,4]])

Seealso:mat

broadcast()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4]])
>>>b=array([5,6])
>>>c=broadcast(a,b)
>>>c.nd#thenumberofdimensionsinthebroadcastedresult
2
>>>c.shape#theshapeofthebroadcastedresult
(2,2)
>>>c.size#totalsizeofthebroadcastedresult
4
>>>forvalueinc:printvalue
...
(1,5)
(2,6)
(3,5)
(4,6)
>>>c.reset()#resettheiteratortothebeginning
>>>c.next()#nextelement
(1,5)

Seealso:ndenumerate,ndindex,flat

bytes()
>>>fromnumpyimport*
>>>fromnumpy.randomimportbytes
>>>printrepr(bytes(5))#stringof5randombytes
'o\x07\x9f\xdf\xdf'
>>>printrepr(bytes(5))#anotherstringof5randombytes
'\x98\xc9KD\xe0'

Seealso:shuffle,permutation,seed

c_[]
>>>fromnumpyimport*
>>>c_[1:5]#forsingleranges,c_worksliker_
array([1,2,3,4])
>>>c_[1:5,2:6]#forcommaseparatedvalues,c_stackscolumn

wise
array([[1,2],
[2,3],
[3,4],
[4,5]])
>>>a=array([[1,2,3],[4,5,6]])
>>>c_[a,a]#concatenationalonglast(default)axis(column
wise,that'swhyit'scalledc_)
array([[1,2,3,1,2,3],
[4,5,6,4,5,6]])
>>>c_['0',a,a]#concatenationalong1staxis,equivalentto
r_[a,a]
array([[1,2,3],
[4,5,6],
[1,2,3],
[4,5,6]])

Seealso:r_,hstack,vstack,column_stack,concatenate,bmat,s_

cast[]()
>>>fromnumpyimport*
>>>x=arange(3)
>>>x.dtype
dtype('int32')
>>>cast['int64'](x)
array([0,1,2],dtype=int64)
>>>cast['uint'](x)
array([0,1,2],dtype=uint32)
>>>cast[float128](x)
array([0.0,1.0,2.0],dtype=float128)
>>>cast.keys()#listdtypecastpossibilities
<snip>

Seealso:astype,typeDict

ceil()
>>>fromnumpyimport*
>>>a=array([1.7,1.5,0.2,0.2,1.5,1.7])
>>>ceil(a)#nearestintegersgreaterthanorequaltoa
array([1.,1.,0.,1.,2.,2.])

Seealso:floor,round_,fix,astype

choose()
>>>fromnumpyimport*
>>>choice0=array([10,12,14,16])#selectorandchoicearrays

mustbeequallysized
>>>choice1=array([20,22,24,26])
>>>choice2=array([30,32,34,36])
>>>selector=array([0,0,2,1])#selectorcanonlycontain
integersinrange(number_of_choice_arrays)
>>>selector.choose(choice0,choice1,choice2)
array([10,12,34,26])
>>>a=arange(4)
>>>choose(a>=2,(choice0,choice1))#separatefunctionalso
exists
array([10,12,24,26])

Seealso:compress,take,where,select

clip()
>>>fromnumpyimport*
>>>a=array([5,15,25,3,13])
>>>a.clip(min=10,max=20)
array([10,15,20,10,13])
>>>clip(a,10,20)#thissyntaxalsoexists

Seealso:wherecompress

column_stack()
>>>fromnumpyimport*
>>>a=array([1,2])
>>>b=array([3,4])
>>>c=array([5,6])
>>>column_stack((a,b,c))#a,b,care1darrayswithequallength
array([[1,3,5],
[2,4,6]])

Seealso:concatenate,dstack,hstack,vstack,c_

compress()
>>>fromnumpyimport*
>>>a=array([10,20,30,40])
>>>condition=(a>15)&(a<35)
>>>condition
array([False,True,True,False],dtype=bool)
>>>a.compress(condition)
array([20,30])
>>>a[condition]#sameeffect
array([20,30])
>>>compress(a>=30,a)#thisformalsoexists
array([30,40])

>>>b=array([[10,20,30],[40,50,60]])
>>>b.compress(b.ravel()>=22)
array([30,40,50,60])
>>>x=array([3,1,2])
>>>y=array([50,101])
>>>b.compress(x>=2,axis=1)#illustratestheuseoftheaxis
keyword
array([[10,30],
[40,60]])
>>>b.compress(y>=100,axis=0)
array([[40,50,60]])

Seealso:choose,take,where,trim_zeros,unique,unique1d

concatenate()
>>>fromnumpyimport*
>>>x=array([[1,2],[3,4]])
>>>y=array([[5,6],[7,8]])
>>>concatenate((x,y))#defaultisaxis=0
array([[1,2],
[3,4],
[5,6],
[7,8]])
>>>concatenate((x,y),axis=1)
array([[1,2,5,6],
[3,4,7,8]])

Seealso:append,column_stack,dstack,hstack,vstack,array_split

conj()
Synonymforconjugate()
Seeconjugate()

conjugate()
>>>a=array([1+2j,34j])
>>>a.conj()#.conj()and.conjugate()arethesame
array([1.2.j,3.+4.j])
>>>a.conjugate()
array([1.2.j,3.+4.j])
>>>conj(a)#isalsopossible
>>>conjugate(a)#isalsopossible

Seealso:vdot

copy()

>>>fromnumpyimport*
>>>a=array([1,2,3])
>>>a
array([1,2,3])
>>>b=a#bisareferencetoa
>>>b[1]=4
>>>a
array([1,4,3])
>>>a=array([1,2,3])
>>>b=a.copy()#bisnowanindependentcopyofa
>>>b[1]=4
>>>a
array([1,2,3])
>>>b
array([1,4,3])

Seealso:view

corrcoef()
>>>fromnumpyimport*
>>>T=array([1.3,4.5,2.8,3.9])#temperaturemeasurements
>>>P=array([2.7,8.7,4.7,8.2])#correspondingpressure
measurements
>>>printcorrcoef([T,P])#correlationmatrixoftemperatureand
pressure
[[1.0.98062258]
[0.980622581.]]
>>>rho=array([8.5,5.2,6.9,6.5])#correspondingdensity
measurements
>>>data=column_stack([T,P,rho])
>>>printcorrcoef([T,P,rho])#correlationmatrixofT,Pandrho
[[1.0.980622580.97090288]
[0.980622581.0.91538464]
[0.970902880.915384641.]]

Seealso:cov,var

cos()
>>>cos(array([0,pi/2,pi]))
array([1.00000000e+00,6.12303177e17,1.00000000e+00])

cov()
>>>fromnumpyimport*
>>>x=array([1.,3.,8.,9.])

>>>variance=cov(x)#normalizedbyN1
>>>variance=cov(x,bias=1)#normalizedbyN
>>>T=array([1.3,4.5,2.8,3.9])#temperaturemeasurements
>>>P=array([2.7,8.7,4.7,8.2])#correspondingpressure
measurements
>>>cov(T,P)#covariancebetweentemperatureandpressure
3.9541666666666657
>>>rho=array([8.5,5.2,6.9,6.5])#correspondingdensity
measurements
>>>data=column_stack([T,P,rho])
>>>printcov(data)#covariancematrixofT,Pandrho
[[1.975833333.954166671.85583333]
[3.954166678.229166673.57083333]
[1.855833333.570833331.84916667]]

Seealso:corrcoef,std,var

cross()
>>>fromnumpyimport*
>>>x=array([1,2,3])
>>>y=array([4,5,6])
>>>cross(x,y)#vectorcrossproduct
array([3,6,3])

Seealso:inner,ix_,outer

cumprod()
>>>fromnumpyimport*
>>>a=array([1,2,3])
>>>a.cumprod()#totalproduct1*2*3=6,andintermediate
results1,1*2
array([1,2,6])
>>>cumprod(a)#alsoexists
array([1,2,6])
>>>a=array([[1,2,3],[4,5,6]])
>>>a.cumprod(dtype=float)#specifytypeofoutput
array([1.,2.,6.,24.,120.,720.])
>>>a.cumprod(axis=0)#foreachofthe3columns:productand
intermediateresults
array([[1,2,3],
[4,10,18]])
>>>a.cumprod(axis=1)#foreachofthetworows:productand
intermediateresults
array([[1,2,6],
[4,20,120]])

Seealso:accumulate,prod,cumsum

cumsum()
>>>fromnumpyimport*
>>>a=array([1,2,3])#cumulativesum=intermediatesumming
results&totalsum
>>>a.cumsum()
array([1,3,6])
>>>cumsum(a)#alsoexists
array([1,3,6])
>>>a=array([[1,2,3],[4,5,6]])
>>>a.cumsum(dtype=float)#specifiestypeofoutputvalue(s)
array([1.,3.,6.,10.,15.,21.])
>>>a.cumsum(axis=0)#sumoverrowsforeachofthe3columns
array([[1,2,3],
[5,7,9]])
>>>a.cumsum(axis=1)#sumovercolumnsforeachofthe2rows
array([[1,3,6],
[4,9,15]])

Seealso:accumulate,sum,cumprod

delete()
>>>fromnumpyimport*
>>>a=array([0,10,20,30,40])
>>>delete(a,[2,4])#removea[2]anda[4]
array([0,10,30])
>>>a=arange(16).reshape(4,4)
>>>a
array([[0,1,2,3],
[4,5,6,7],
[8,9,10,11],
[12,13,14,15]])
>>>delete(a,s_[1:3],axis=0)#removerows1and2
array([[0,1,2,3],
[12,13,14,15]])
>>>delete(a,s_[1:3],axis=1)#removecolumns1and2
array([[0,3],
[4,7],
[8,11],
[12,15]])

Seealso:append,insert

det()
>>>fromnumpyimport*
>>>fromnumpy.linalgimportdet
>>>A=array([[1.,2.],[3.,4.]])

>>>det(A)#determinantofsquarematrix
2.0

Seealso:inv

diag()
>>>fromnumpyimport*
>>>a=arange(12).reshape(4,3)
>>>printa
[[012]
[345]
[678]
[91011]]
>>>printdiag(a,k=0)
[048]
>>>printdiag(a,k=1)
[15]
>>>printdiag(array([1,4,5]),k=0)
[[100]
[040]
[005]]
>>>printdiag(array([1,4,5]),k=1)
[[0100]
[0040]
[0005]
[0000]]

Seealso:diagonal,diagflat,trace

diagflat()
>>>fromnumpyimport*
>>>x=array([[5,6],[7,8]])
>>>diagflat(x)#flattenx,thenputelementsondiagonal
array([[5,0,0,0],
[0,6,0,0],
[0,0,7,0],
[0,0,0,8]])

Seealso:diag,diagonal,flatten

diagonal()
>>>fromnumpyimport*
>>>a=arange(12).reshape(3,4)
>>>printa
[[0123]
[4567]

[891011]]
>>>a.diagonal()
array([0,5,10])
>>>a.diagonal(offset=1)
array([1,6,11])
>>>diagonal(a)#Alsothisformexists
array([0,5,10])

Seealso:diag,diagflat,trace

diff()
>>>fromnumpyimport*
>>>x=array([0,1,3,9,5,10])
>>>diff(x)#1storderdifferencesbetweentheelementsofx
array([1,2,6,4,5])
>>>diff(x,n=2)#2ndorderdifferences,equivalentto
diff(diff(x))
array([1,4,10,9])
>>>x=array([[1,3,6,10],[0,5,6,8]])
>>>diff(x)#1storderdifferencesbetweenthecolumns(default:
axis=1)
array([[2,3,4],
[5,1,2]])
>>>diff(x,axis=0)#1storderdifferencebetweentherows
array([[1,2,0,2]])

digitize()
>>>fromnumpyimport*
>>>x=array([0.2,6.4,3.0,1.6])
>>>bins=array([0.0,1.0,2.5,4.0,10.0])#monotonically
increasing
>>>d=digitize(x,bins)#inwhichbinfallseachvalueofx?
>>>d
array([1,4,3,2])
>>>forninrange(len(x)):
...printbins[d[n]1],"<=",x[n],"<",bins[d[n]]
...
0.0<=0.2<1.0
4.0<=6.4<10.0
2.5<=3.0<4.0
1.0<=1.6<2.5

Seealso:bincount,histogram

dot()

>>>fromnumpyimport*
>>>x=array([[1,2,3],[4,5,6]])
>>>x.shape
(2,3)
>>>y=array([[1,2],[3,4],[5,6]])
>>>y.shape
(3,2)
>>>dot(x,y)#matrixmultiplication(2,3)x(3,2)>(2,2)
array([[22,28],
[49,64]])
>>>
>>>importnumpy
>>>ifid(dot)==id(numpy.core.multiarray.dot):#Awaytoknow
ifyouusefastblas/lapackornot.
...print"Notusingblas/lapack!"

Seealso:vdot,inner,multiply

dsplit()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4]])
>>>b=dstack((a,a,a,a))
>>>b.shape#stackingindepth:forkin(0,..,3):b[:,:,k]=a
(2,2,4)
>>>c=dsplit(b,2)#split,depthwise,in2equalparts
>>>printc[0].shape,c[1].shape#forkin(0,1):c[0][:,:,k]=a
andc[1][:,:,k]=a
(2,2,2)(2,2,2)
>>>d=dsplit(b,[1,2])#splitbefore[:,:,1]andbefore[:,:,2]
>>>printd[0].shape,d[1].shape,d[2].shape#foranyofthe
parts:d[.][:,:,k]=a
(2,2,1)(2,2,1)(2,2,2)

Seealso:split,array_split,hsplit,vsplit,dstack

dstack()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4]])#shapesofaandbcanonlydiffer
inthe3rddimension(ifpresent)
>>>b=array([[5,6],[7,8]])
>>>dstack((a,b))#stackarraysalongathirdaxis(depthwise)
array([[[1,5],
[2,6]],
[[3,7],
[4,8]]])

Seealso:column_stack,concatenate,hstack,vstack,dsplit

dtype()
>>>fromnumpyimport*
>>>dtype('int16')#usingarrayscalartype
dtype('int16')
>>>dtype([('f1','int16')])#record,1fieldnamed'f1',
containingint16
dtype([('f1','<i2')])
>>>dtype([('f1',[('f1','int16')])])#record,1fieldnamed
'f1'containingarecordthathas1field.
dtype([('f1',[('f1','<i2')])])
>>>dtype([('f1','uint'),('f2','int32')])#recordwith2
fields:field1containsanunsignedint,2ndfieldanint32
dtype([('f1','<u4'),('f2','<i4')])
>>>dtype([('a','f8'),('b','S10')])#usingarrayprotocoltype
strings
dtype([('a','<f8'),('b','|S10')])
>>>dtype("i4,(2,3)f8")#usingcommaseparatedfieldformats.
(2,3)istheshape
dtype([('f0','<i4'),('f1','<f8',(2,3))])
>>>dtype([('hello',('int',3)),('world','void',10)])#using
tuples.intisfixedtype:3isshapevoidisflextype:10is
size.
dtype([('hello','<i4',3),('world','|V10')])
>>>dtype(('int16',{'x':('int8',0),'y':('int8',1)}))#subdivide
int16in2int8,calledxandy.0and1aretheoffsetsinbytes
dtype(('<i2',[('x','|i1'),('y','|i1')]))
>>>dtype({'names':['gender','age'],'formats':['S1',uint8]})#
usingdictionaries.2fieldsnamed'gender'and'age'
dtype([('gender','|S1'),('age','|u1')])
>>>dtype({'surname':('S25',0),'age':(uint8,25)})#0and25are
offsetsinbytes
dtype([('surname','|S25'),('age','|u1')])
>>>
>>>a=dtype('int32')
>>>a
dtype('int32')
>>>a.type#typeobject
<type'numpy.int32'>
>>>a.kind#charactercode(oneof'biufcSUV')toidentify
generaltype
'i'
>>>a.char#uniquecharcodeofeachofthe21builtintypes
'l'
>>>a.num#uniquenumberofeachofthe21builtintypes
7
>>>a.str#arrayprotocoltypestring
'<i4'
>>>a.name#nameofthisdatatype
'int32'

>>>a.byteorder#'=':native,'<':littleendian,'>':bigendian,
'|':notapplicable
'='
>>>a.itemsize#itemsizeinbytes
4
>>>a=dtype({'surname':('S25',0),'age':(uint8,25)})
>>>a.fields.keys()
['age','surname']
>>>a.fields.values()
[(dtype('uint8'),25),(dtype('|S25'),0)]
>>>a=dtype([('x','f4'),('y','f4'),#nestedfield
...('nested',[('i','i2'),('j','i2')])])
>>>a.fields['nested']#accessnestedfields
(dtype([('i','<i2'),('j','<i2')]),8)
>>>a.fields['nested'][0].fields['i']#accessnestedfields
(dtype('int16'),0)
>>>a.fields['nested'][0].fields['i'][0].type
<type'numpy.int16'>

Seealso:array,typeDict,astype

empty()
>>>fromnumpyimport*
>>>empty(3)#uninitializedarray,size=3,dtype=float
array([6.08581638e+000,3.45845952e323,4.94065646e324])
>>>empty((2,3),int)#uninitializedarray,dtype=int
array([[1075337192,1075337192,135609024],
[1084062604,1197436517,1129066306]])

Seealso:ones,zeros,eye,identity

empty_like()
>>>fromnumpyimport*
>>>a=array([[1,2,3],[4,5,6]])
>>>empty_like(a)#uninitializedarraywiththesameshapeand
datatypeas'a'
array([[0,25362433,6571520],
[21248,136447968,4]])

Seealso:ones_like,zeros_like

expand_dims()
>>>fromnumpyimport*
>>>x=array([1,2])
>>>expand_dims(x,axis=0)#Equivalenttox[newaxis,:]orx[None]
orx[newaxis]

array([[1,2]])
>>>expand_dims(x,axis=1)#Equivalenttox[:,newaxis]
array([[1],
[2]])

Seealso:newaxis,atleast_1d,atleast_2d,atleast_3d

eye()
>>>fromnumpyimport*
>>>eye(3,4,0,dtype=float)#a3x4matrixcontainingzerosexcept
forthe0thdiagonalthatcontainsones
array([[1.,0.,0.,0.],
[0.,1.,0.,0.],
[0.,0.,1.,0.]])
>>>eye(3,4,1,dtype=float)#a3x4matrixcontainingzerosexcept
forthe1stdiagonalthatcontainsones
array([[0.,1.,0.,0.],
[0.,0.,1.,0.],
[0.,0.,0.,1.]])

Seealso:ones,zeros,empty,identity

fft()
>>>fromnumpyimport*
>>>fromnumpy.fftimport*
>>>signal=array([2.,8.,6.,4.,1.,0.,3.,5.])#could
alsobecomplex
>>>fourier=fft(signal)
>>>fourier
array([13.+0.j,3.36396103+4.05025253j,
2.+1.j,9.3639610313.94974747j,
21.+0.j,9.36396103+13.94974747j,
2.1.j,3.363961034.05025253j])
>>>
>>>N=len(signal)
>>>fourier=empty(N,complex)
>>>forkinrange(N):#equivalentbutmuchslower
...fourier[k]=sum(signal*exp(1j*2*pi*k*arange(N)/N))
...
>>>timestep=0.1#ifunit=day>frequnit=cycles/day
>>>fftfreq(N,d=timestep)#freqscorrespondingto'fourier'
array([0.,1.25,2.5,3.75,5.,3.75,2.5,1.25])

Seealso:ifft,fftfreq,fftshift

fftfreq()

>>>fromnumpyimport*
>>>fromnumpy.fftimport*
>>>signal=array([2.,8.,6.,4.,1.,0.,3.,5.])
>>>fourier=fft(signal)
>>>N=len(signal)
>>>timestep=0.1#ifunit=day>frequnit=cycles/day
>>>freq=fftfreq(N,d=timestep)#freqscorrespondingto
'fourier'
>>>freq
array([0.,1.25,2.5,3.75,5.,3.75,2.5,1.25])
>>>
>>>fftshift(freq)#freqsinascendingorder
array([5.,3.75,2.5,1.25,0.,1.25,2.5,3.75])

Seealso:fft,ifft,fftshift

fftshift()
>>>fromnumpyimport*
>>>fromnumpy.fftimport*
>>>signal=array([2.,8.,6.,4.,1.,0.,3.,5.])
>>>fourier=fft(signal)
>>>N=len(signal)
>>>timestep=0.1#ifunit=day>frequnit=cycles/day
>>>freq=fftfreq(N,d=timestep)#freqscorrespondingto
'fourier'
>>>freq
array([0.,1.25,2.5,3.75,5.,3.75,2.5,1.25])
>>>
>>>freq=fftshift(freq)#freqsinascendingorder
>>>freq
array([5.,3.75,2.5,1.25,0.,1.25,2.5,3.75])
>>>fourier=fftshift(fourier)#adjustfouriertonewfreqorder
>>>
>>>freq=ifftshift(freq)#undopreviousfrequencyshift
>>>fourier=ifftshift(fourier)#undopreviousfouriershift

Seealso:fft,ifft,fftfreq

fill()
>>>fromnumpyimport*
>>>a=arange(4,dtype=int)
>>>a
array([0,1,2,3])
>>>a.fill(7)#replaceallelementswiththenumber7
>>>a
array([7,7,7,7])
>>>a.fill(6.5)#fillvalueisconvertedtodtypeofa

>>>a
array([6,6,6,6])

Seealso:empty,zeros,ones,repeat

finfo()
>>>fromnumpyimport*
>>>f=finfo(float)#thenumbersgivenaremachinedependent
>>>f.nmant,f.nexp#nrofbitsinthemantissaandinthe
exponent
(52,11)
>>>f.machep#mostnegativensothat1.0+2**n!=1.0
52
>>>f.eps#floatingpointprecision:2**machep
array(2.2204460492503131e16)
>>>f.precision#nrofprecisedecimaldigits:int(log10(eps))
15
>>>f.resolution#10**(precision)
array(1.0000000000000001e15)
>>>f.negep#mostnegativensothat1.02**n!=1.0
53
>>>f.epsneg#floatingpointprecision:2**negep
array(1.1102230246251565e16)
>>>f.minexp#mostnegativensothat2**ngivesnormalnumbers
1022
>>>f.tiny#smallestusuablefloatingpointnr:2**minexp
array(2.2250738585072014e308)
>>>f.maxexp#smallestpositivensothat2**ncausesoverflow
1024
>>>f.min,f.max#themostnegativeandmostpositiveusuable
floatingnumber
(1.7976931348623157e+308,array(1.7976931348623157e+308))

fix()
>>>fromnumpyimport*
>>>a=array([1.7,1.5,0.2,0.2,1.5,1.7])
>>>fix(a)#roundatonearestintegertowardszero
array([1.,1.,0.,0.,1.,1.])

Seealso:round_,ceil,floor,astype

flat
>>>fromnumpyimport*
>>>a=array([[10,30],[40,60]])
>>>iter=a.flat#.flatreturnsaniterator

>>>iter.next()#cyclethrougharraywith.next()
10
>>>iter.next()
30
>>>iter.next()
40

Seealso:broadcast,flatten

flatten()
>>>fromnumpyimport*
>>>a=array([[[1,2]],[[3,4]]])
>>>printa
[[[12]]
[[34]]]
>>>b=a.flatten()#bisnowa1dversionofa,anewarray,
notareference
>>>printb
[1234]

Seealso:ravel,flat

fliplr()
>>>fromnumpyimport*
>>>a=arange(12).reshape(4,3)
>>>a
array([[0,1,2],
[3,4,5],
[6,7,8],
[9,10,11]])
>>>fliplr(a)#flipleftright
array([[2,1,0],
[5,4,3],
[8,7,6],
[11,10,9]])

Seealso:flipud,rot90

flipud()
>>>fromnumpyimport*
>>>a=arange(12).reshape(4,3)
>>>a
array([[0,1,2],
[3,4,5],
[6,7,8],
[9,10,11]])

>>>flipud(a)#flipupdown
array([[9,10,11],
[6,7,8],
[3,4,5],
[0,1,2]])

Seealso:fliplr,rot90

floor()
>>>fromnumpyimport*
>>>a=array([1.7,1.5,0.2,0.2,1.5,1.7])
>>>floor(a)
array([2.,2.,1.,0.,1.,1.])#nearestintegersmallerthan
orequaltoa#nearestintegersgreaterthanorequaltoa

Seealso:ceil,round_,fix,astype

fromarrays()
>>>fromnumpyimport*
>>>x=array(['Smith','Johnson','McDonald'])#datatypeisstring
>>>y=array(['F','F','M'],dtype='S1')#datatypeisasingle
character
>>>z=array([20,25,23])#datatypeisinteger
>>>data=rec.fromarrays([x,y,z],names='surname,gender,age')#
converttorecordarray
>>>data[0]
('Smith','F',20)
>>>data.age#namesareavailableasattributes
array([20,25,23])

Seealso:view

frombuffer()
>>>fromnumpyimport*
>>>buffer="\x00\x00\x00\x00\x00\x00\xf0?
\x00\x00\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00\x00\x08\
...
@\x00\x00\x00\x00\x00\x00\x10@\x00\x00\x00\x00\x00\x00\x14@\x00\x0
0\x00\x00\x00\x00\x18@"
>>>a=frombuffer(buffer,complex128)
>>>a
array([1.+2.j,3.+4.j,5.+6.j])

Seealso:fromfunction,fromfile

fromfile()

>>>fromnumpyimport*
>>>y=array([2.,4.,6.,8.])
>>>y.tofile("myfile.dat")#binaryformat
>>>y.tofile("myfile.txt",sep='\n',format="%e")#ascii
format,onecolumn,exponentialnotation
>>>fromfile('myfile.dat',dtype=float)
array([2.,4.,6.,8.])
>>>fromfile('myfile.txt',dtype=float,sep='\n')
array([2.,4.,6.,8.])

Seealso:loadtxt,fromfunction,tofile,frombuffer,savetxt

fromfunction()
>>>fromnumpyimport*
>>>deff(i,j):
...returni**2+j**2
...
>>>fromfunction(f,(3,3))#evaluatefunctiomforall
combinationsofindices[0,1,2]x[0,1,2]
array([[0,1,4],
[1,2,5],
[4,5,8]])

Seealso:fromfile,frombuffer

fromiter()

>>>fromnumpyimport*
>>>importitertools
>>>mydata=[[55.5,40],[60.5,70]]#Listoflists
>>>mydescriptor={'names':('weight','age'),'formats':
(float32,int32)}#Descriptorofthedata
>>>myiterator=itertools.imap(tuple,mydata)#Cleverwayof
puttinglistoflistsintoiterator

#oftuples.E.g.:myiterator.next()==(55.5,40.)
>>>a=fromiter(myiterator,dtype=mydescriptor)
>>>a
array([(55.5,40),(60.5,70)],
dtype=[('weight','<f4'),('age','<i4')])

Seealso:fromarrays,frombuffer,fromfile,fromfunction

generic
>>>fromnumpyimport*

>>>numpyscalar=string_('7')#Converttonumpyscalar
>>>numpyscalar#Lookslikeabuildinscalar...
'7'
>>>type(numpyscalar)#...butitisn't
<type'numpy.string_'>
>>>buildinscalar='7'#Buildinpythonscalar
>>>type(buildinscalar)
<type'str'>
>>>isinstance(numpyscalar,generic)#CheckifscalarisaNumPy
one
True
>>>isinstance(buildinscalar,generic)#Exampleonhowto
recognizeNumPyscalars
False

gumbel()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>gumbel(loc=0.0,scale=1.0,size=(2,3))#Gumbeldistribution
location=0.0,scale=1.0
array([[1.25923601,1.68758144,1.76620507],
[1.96820048,0.21219499,1.83579566]])
>>>frompylabimport*#histogramplotexample
>>>hist(gumbel(0,1,(1000)),50)

Seealso:random_sample,uniform,poisson,seed

histogram()
>>>fromnumpyimport*
>>>x=array([0.2,6.4,3.0,1.6,0.9,2.3,1.6,5.7,8.5,4.0,
12.8])
>>>bins=array([0.0,1.0,2.5,4.0,10.0])#increasing
monotonically
>>>N,bins=histogram(x,bins)
>>>N,bins
(array([2,3,1,4]),array([0.,1.,2.5,4.,10.]))
>>>forninrange(len(bins)1):
...print"#",N[n],"numberfallintobin[",bins[n],",",
bins[n+1],"["
...
#2numbersfallintobin[0.0,1.0[
#3numbersfallintobin[1.0,2.5[
#1numbersfallintobin[2.5,4.0[
#4numbersfallintobin[4.0,10.0[
#
>>>N,bins=histogram(x,5,range=(0.0,10.0))#5binboundaries
intherange(0,10)

>>>N,bins
(array([4,2,2,1,2]),array([0.,2.,4.,6.,8.]))
>>>N,bins=histogram(x,5,range=(0.0,10.0),normed=True)#
normalizehistogram,i.e.dividebylen(x)
>>>N,bins
(array([0.18181818,0.09090909,0.09090909,0.04545455,
0.09090909]),array([0.,2.,4.,6.,8.]))

Seealso:bincount,digitize

hsplit()
>>>fromnumpyimport*
>>>a=array([[1,2,3,4],[5,6,7,8]])
>>>hsplit(a,2)#split,columnwise,in2equalparts
[array([[1,2],
[5,6]]),array([[3,4],
[7,8]])]
>>>hsplit(a,[1,2])#splitbeforecolumn1andbeforecolumn2
[array([[1],
[5]]),array([[2],
[6]]),array([[3,4],
[7,8]])]

Seealso:split,array_split,dsplit,vsplit,hstack

hstack()
>>>fromnumpyimport*
>>>a=array([[1],[2]])#2x1array
>>>b=array([[3,4],[5,6]])#2x2array
>>>hstack((a,b,a))#onlythe2nddimensionofthearraysis
allowedtobedifferent
array([[1,3,4,1],
[2,5,6,2]])

Seealso:column_stack,concatenate,dstack,vstack,hsplit

hypot()
>>>fromnumpyimport*
>>>hypot(3.,4.)#hypotenuse:sqrt(3**2+4**2)=5
5.0
>>>z=array([2+3j,3+4j])
>>>hypot(z.real,z.imag)#normofcomplexnumbers
array([3.60555128,5.])

Seealso:angle,abs

identity()
>>>fromnumpyimport*
>>>identity(3,float)
array([[1.,0.,0.],
[0.,1.,0.],
[0.,0.,1.]])

Seealso:empty,eye,ones,zeros

ifft()
>>>fromnumpyimport*
>>>fromnumpy.fftimport*
>>>signal=array([2.,8.,6.,4.,1.,0.,3.,5.])
>>>fourier=fft(signal)
>>>ifft(fourier)#Inversefouriertransform
array([2.+0.00000000e+00j,8.+1.51410866e15j,6.
+3.77475828e15j,
4.+2.06737026e16j,1.+0.00000000e+00j,0.1.92758271e
15j,
3.3.77475828e15j,5.+2.06737026e16j])
>>>
>>>allclose(signal.astype(complex),ifft(fft(signal)))#
ifft(fft())=originalsignal
True
>>>
>>>N=len(fourier)
>>>signal=empty(N,complex)
>>>forkinrange(N):#equivalentbutmuchslower
...signal[k]=sum(fourier*exp(+1j*2*pi*k*arange(N)/N))/N

Seealso:fft,fftfreq,fftshift

imag
>>>fromnumpyimport*
>>>a=array([1+2j,3+4j,5+6j])
>>>a.imag
array([2.,4.,6.])
>>>a.imag=9
>>>a
array([1.+9.j,3.+9.j,5.+9.j])
>>>a.imag=array([9,8,7])
>>>a
array([1.+9.j,3.+8.j,5.+7.j])

Seealso:real,angle

index_exp[]
>>>fromnumpyimport*
>>>myslice=index_exp[2:4,...,4,::1]#myslicecouldnowbe
passedtoafunction,forexample.
>>>printmyslice
(slice(2,4,None),Ellipsis,4,slice(None,None,1))

Seealso:slice,s_

indices()
>>>fromnumpyimport*
>>>indices((2,3))
array([[[0,0,0],
[1,1,1]],
[[0,1,2],
[0,1,2]]])
>>>a=array([[0,1,2,3,4],
...[10,11,12,13,14],
...[20,21,22,23,24],
...[30,31,32,33,34]])
>>>i,j=indices((2,3))
>>>a[i,j]
array([[0,1,2],
[10,11,12]])

Seealso:mgrid,[],ix_,slice

inf
>>>fromnumpyimport*
>>>exp(array([1000.]))#inf=infinite=numbertoolargeto
represent,machinedependent
array([inf])
>>>x=array([2,inf,1,inf])
>>>isfinite(x)#showwhichelementsarenotnan/inf/inf
array([True,False,True,False],dtype=bool)
>>>isinf(x)#showwhichelementsareinf/inf
array([False,True,False,True],dtype=bool)
>>>isposinf(x)#showwhichelementsareinf
array([False,False,False,True],dtype=bool)
>>>isneginf(x)#showwhichelementsareinf
array([False,True,False,False],dtype=bool)
>>>nan_to_num(x)#replaceinf/infwithmostnegative/positive
representablenumber
array([2.00000000e+000,1.79769313e+308,1.00000000e+000,
1.79769313e+308])

Seealso:nan,finfo

inner()
>>>fromnumpyimport*
>>>x=array([1,2,3])
>>>y=array([10,20,30])
>>>inner(x,y)#1x10+2x20+3x30=140
140

Seealso:cross,outer,dot

insert()
>>>fromnumpyimport*
>>>a=array([10,20,30,40])
>>>insert(a,[1,3],50)#insertvalue50beforeelements[1]and
[3]
array([10,50,20,30,50,40])
>>>insert(a,[1,3],[50,60])#insertvalue50beforeelement[1]
andvalue60beforeelement[3]
array([10,50,20,30,60,40])
>>>a=array([[10,20,30],[40,50,60],[70,80,90]])
>>>insert(a,[1,2],100,axis=0)#insertrowwithvalues100
beforerow[1]andbeforerow[2]
array([[10,20,30],
[100,100,100],
[40,50,60],
[100,100,100],
[70,80,90]])
>>>insert(a,[0,1],[[100],[200]],axis=0)
array([[100,100,100],
[10,20,30],
[200,200,200],
[40,50,60],
[70,80,90]])
>>>insert(a,[0,1],[100,200],axis=1)
array([[100,10,200,20,30],
[100,40,200,50,60],
[100,70,200,80,90]])

Seealso:delete,append

inv()
>>>fromnumpyimport*
>>>fromnumpy.linalgimportinv
>>>a=array([[3,1,5],[1,0,8],[2,1,4]])
>>>printa

[[315]
[108]
[214]]
>>>inva=inv(a)#Inversematrix
>>>printinva
[[1.142857140.142857141.14285714]
[1.714285710.285714292.71428571]
[0.142857140.142857140.14285714]]
>>>dot(a,inva)#Checktheresult,shouldbeeye(3)within
machineprecision
array([[1.00000000e00,2.77555756e17,3.60822483e16],
[0.00000000e+00,1.00000000e+00,0.00000000e+00],
[1.11022302e16,0.00000000e+00,1.00000000e+00]])

Seealso:solve,pinv,det

iscomplex()
>>>importnumpyasnp
>>>a=np.array([1,2,3.j])
>>>np.iscomplex(a)
array([False,False,True],dtype=bool)

iscomplexobj()
>>>importnumpyasnp
>>>a=np.array([1,2,3.j])
>>>np.iscomplexobj(a)
True
>>>a=np.array([1,2,3])
>>>np.iscomplexobj(a)
False
>>>a=np.array([1,2,3],dtype=np.complex)
>>>np.iscomplexobj(a)
True

item()
>>>fromnumpyimport*
>>>a=array([5])
>>>type(a[0])
<type'numpy.int32'>
>>>a.item()#Conversionofarrayofsize1toPythonscalar
5
>>>type(a.item())
<type'int'>
>>>b=array([2,3,4])
>>>b[1].item()#Conversionof2ndelementtoPythonscalar

3
>>>type(b[1].item())
<type'int'>
>>>b.item(2)#Return3rdelementconvertedtoPythonscalar
4
>>>type(b.item(2))
<type'int'>
>>>type(b[2])#b[2]isslowerthanb.item(2),andthereisno
conversion
<type'numpy.int32'>

Seealso:[]

ix_()
>>>fromnumpyimport*
>>>a=arange(9).reshape(3,3)
>>>printa
[[012]
[345]
[678]]
>>>indices=ix_([0,1,2],[1,2,0])#tricktobeusedwitharray
broadcasting
>>>printindices
(array([[0],
[1],
[2]]),array([[1,2,0]]))
>>>printa[indices]
[[120]
[453]
[786]]
>>>#Thelatterarrayisthecrossproduct:
>>>#[[a[0,1]a[0,2]a[0,0]]
...#[a[1,1]a[1,2]a[1,0]]
...#[a[2,1]a[2,2]a[2,0]]]
...

Seealso:[],indices,cross,outer

lexsort()
>>>fromnumpyimport*
>>>serialnr=array([1023,5202,6230,1671,1682,5241])
>>>height=array([40.,42.,60.,60.,98.,40.])
>>>width=array([50.,20.,70.,60.,15.,30.])
>>>
>>>#Wewanttosorttheserialnumberswithincreasingheight,
_AND_
>>>#serialnumberswithequalheightsshouldbesortedwith
increasingwidth.

>>>
>>>indices=lexsort(keys=(width,height))#mindtheorder!
>>>indices
array([5,0,1,3,2,4])
>>>forninindices:
...printserialnr[n],height[n],width[n]
...
524140.030.0
102340.050.0
520242.020.0
167160.060.0
623060.070.0
168298.015.0
>>>
>>>a=vstack([serialnr,width,height])#Alternatively:alldata
inonebigmatrix
>>>printa#Mindtheorderoftherows!
[[1023.5202.6230.1671.1682.5241.]
[50.20.70.60.15.30.]
[40.42.60.60.98.40.]]
>>>indices=lexsort(a)#Sortonlastrow,thenon2ndlastrow,
etc.
>>>a.take(indices,axis=1)
array([[5241.,1023.,5202.,1671.,6230.,1682.],
[30.,50.,20.,60.,70.,15.],
[40.,40.,42.,60.,60.,98.]])

Seealso:sort,argsort

linspace()
>>>fromnumpyimport*
>>>linspace(0,5,num=6)#6evenlyspacednumbersbetween0and5
incl.
array([0.,1.,2.,3.,4.,5.])
>>>linspace(0,5,num=10)#10evenlyspacednumbersbetween0and
5incl.
array([0.,0.55555556,1.11111111,1.66666667,2.22222222,
2.77777778,3.33333333,3.88888889,4.44444444,5.])
>>>linspace(0,5,num=10,endpoint=False)#10evenlyspacednumbers
between0and5EXCL.
array([0.,0.5,1.,1.5,2.,2.5,3.,3.5,4.,4.5])
>>>stepsize=linspace(0,5,num=10,endpoint=False,retstep=True)#
besidestheusualarray,alsoreturnthestepsize
>>>stepsize
(array([0.,0.5,1.,1.5,2.,2.5,3.,3.5,4.,4.5]),0.5)
>>>myarray,stepsize=
linspace(0,5,num=10,endpoint=False,retstep=True)
>>>stepsize
0.5

Seealso:arange,logspace,r_

loadtxt()
>>>fromnumpyimport*
>>>
>>>data=loadtxt("myfile.txt")#myfile.txtcontains4columns
ofnumbers
>>>t,z=data[:,0],data[:,3]#datais2Dnumpyarray
>>>
>>>t,x,y,z=loadtxt("myfile.txt",unpack=True)#tounpackall
columns
>>>t,z=loadtxt("myfile.txt",usecols=(0,3),unpack=True)#to
selectjustafewcolumns
>>>data=loadtxt("myfile.txt",skiprows=7)#toskip7rows
fromtopoffile
>>>data=loadtxt("myfile.txt",comments='!')#use'!'as
commentcharinsteadof'#'
>>>data=loadtxt("myfile.txt",delimiter='')#use''as
columnseparatorinsteadofwhitespace
>>>data=loadtxt("myfile.txt",dtype=int)#filecontains
integersinsteadoffloats

Seealso:savetxt,fromfile

logical_and()
>>>fromnumpyimport*
>>>logical_and(array([0,0,1,1]),array([0,1,0,1]))
array([False,False,False,True],dtype=bool)
>>>logical_and(array([False,False,True,True]),
array([False,True,False,True]))
array([False,False,False,True],dtype=bool)

Seealso:logical_or,logical_not,logical_xor,bitwise_and

logical_not()
>>>fromnumpyimport*
>>>logical_not(array([0,1]))
>>>logical_not(array([False,True]))

Seealso:logical_or,logical_not,logical_xor,bitwise_and

logical_or()
>>>fromnumpyimport*
>>>logical_or(array([0,0,1,1]),array([0,1,0,1]))

>>>logical_or(array([False,False,True,True]),
array([False,True,False,True]))

Seealso:logical_and,logical_not,logical_xor,bitwise_or

logical_xor()
>>>fromnumpyimport*
>>>logical_xor(array([0,0,1,1]),array([0,1,0,1]))
>>>logical_xor(array([False,False,True,True]),
array([False,True,False,True]))

Seealso:logical_or,logical_not,logical_or,bitwise_xor

logspace()
>>>fromnumpyimport*
>>>logspace(2,3,num=6)#6evenlyspacedptsona
logarithmicscale,from10^{2}to10^3incl.
array([1.00000000e02,1.00000000e01,1.00000000e+00,
1.00000000e+01,1.00000000e+02,1.00000000e+03])
>>>logspace(2,3,num=10)#10evenlyspacedptsona
logarithmicscale,from10^{2}to10^3incl.
array([1.00000000e02,3.59381366e02,1.29154967e01,
4.64158883e01,1.66810054e+00,5.99484250e+00,
2.15443469e+01,7.74263683e+01,2.78255940e+02,
1.00000000e+03])
>>>logspace(2,3,num=6,endpoint=False)#6evenlyspacedpts
onalogarithmicscale,from10^{2}to10^3EXCL.
array([1.00000000e02,6.81292069e02,4.64158883e01,
3.16227766e+00,2.15443469e+01,1.46779927e+02])
>>>exp(linspace(log(0.01),log(1000),num=6,endpoint=False))#
forcomparison
array([1.00000000e02,6.81292069e02,4.64158883e01,
3.16227766e+00,2.15443469e+01,1.46779927e+02])

Seealso:arange,linspace,r_

lstsq()
lstsq()ismostoftenusedinthecontextofleastsquaresfittingofdata.Supposeyou
obtainsomenoisydatayasafunctionofavariablet,e.g.velocityasafunctionof
time.Youcanuselstsq()tofitamodeltothedata,ifthemodelislinearinits
parameters,thatisif
y=p0*f0(t)+p1*f1(t)+...+pN1*fN1(t)+noise
wherethepiaretheparametersyouwanttoobtainthroughfittingandthefi(t)are
knownfunctionsoft.Whatfollowsisanexamplehowyoucandothis.
First,fortheexample'ssake,somedataissimulated:

>>>fromnumpyimport*
>>>fromnumpy.randomimportnormal
>>>t=arange(0.0,10.0,0.05)#independentvariable
>>>y=2.0*sin(2.*pi*t*0.6)+2.7*cos(2.*pi*t*0.6)+
normal(0.0,1.0,len(t))

Wewouldliketofitthisdatawith:model(t)=p0*sin(2.*pi*t*0.6)+p1*
cos(2.*pi*t*0.6),wherep0andp1aretheunknownfitparameters.Herewego:
>>>fromnumpy.linalgimportlstsq
>>>Nparam=2#wewanttoestimate2parameters:p_0andp_1
>>>A=zeros((len(t),Nparam),float)#onebigarraywithallthe
f_i(t)
>>>A[:,0]=sin(2.*pi*t*0.6)#f_0(t)stored
>>>A[:,1]=cos(2.*pi*t*0.6)#f_1(t)stored
>>>(p,residuals,rank,s)=lstsq(A,y)
>>>p#ourfinalestimateoftheparametersusingnoisydata
array([1.9315685,2.71165171])
>>>residuals#sumoftheresiduals:sum((p[0]*A[:,0]+p[1]*
A[:,1]y)**2)
array([217.23783374])
>>>rank#rankofthearrayA
2
>>>s#singularvaluesofA
array([10.,10.])

Seealso:pinv,polyfit,solve

mat()
>>>fromnumpyimport*
>>>mat('134569')#matricesarealways2dimensional
matrix([[1,3,4],
[5,6,9]])
>>>a=array([[1,2],[3,4]])
>>>m=mat(a)#convert2darraytomatrix
>>>m
matrix([[1,2],
[3,4]])
>>>a[0]#resultis1dimensional
array([1,2])
>>>m[0]#resultis2dimensional
matrix([[1,2]])
>>>a.ravel()#resultis1dimensional
array([1,2,3,4])
>>>m.ravel()#resultis2dimensional
matrix([[1,2,3,4]])
>>>a*a#elementbyelementmultiplication
array([[1,4],

[9,16]])
>>>m*m#(algebraic)matrixmultiplication
matrix([[7,10],
[15,22]])
>>>a**3#elementwisepower
array([[1,8],
[27,64]])
>>>m**3#matrixmultiplicationm*m*m
matrix([[37,54],
[81,118]])
>>>m.T#transposeofthematrix
matrix([[1,3],
[2,4]])
>>>m.H#conjugatetranspose(differsfrom.Tforcomplex
matrices)
matrix([[1,3],
[2,4]])
>>>m.I#inversematrix
matrix([[2.,1.],
[1.5,0.5]])

Seealso:bmat,array,dot,asmatrix

matrix()
>>>fromnumpyimport*
>>>matrix('134569')#matrixissynonymouswithmat
matrix([[1,3,4],
[5,6,9]])

Seealso:mat,asmatrix

max()
>>>fromnumpyimport*
>>>a=array([10,20,30])
>>>a.max()
30
>>>a=array([[10,50,30],[60,20,40]])
>>>a.max()
60
>>>a.max(axis=0)#foreachofthecolumns,findthemaximum
array([60,50,40])
>>>a.max(axis=1)#foreachoftherows,findthemaximum
array([50,60])
>>>max(a)#alsoexists,butisslower

Seealso:nan,argmax,maximum,ptp

maximum()
>>>fromnumpyimport*
>>>a=array([1,0,5])
>>>b=array([3,2,4])
>>>maximum(a,b)#elementbyelementcomparison
array([3,2,5])
>>>max(a.tolist(),b.tolist())#standardPythonfunctiondoesnot
givethesame!
[3,2,4]

Seealso:minimum,max,argmax

mean()
>>>fromnumpyimport*
>>>a=array([1,2,7])
>>>a.mean()
3.3333333333333335
>>>a=array([[1,2,7],[4,9,6]])
>>>a.mean()
4.833333333333333
>>>a.mean(axis=0)#themeanofeachofthe3columns
array([2.5,5.5,6.5])
>>>a.mean(axis=1)#themeanofeachofthe2rows
array([3.33333333,6.33333333])

Seealso:average,median,var,std,sum

median()
>>>fromnumpyimport*
>>>a=array([1,2,3,4,9])
>>>median(a)
3
>>>a=array([1,2,3,4,9,0])
>>>median(a)
2.5

Seealso:average,mean,var,std

mgrid[]
>>>fromnumpyimport*
>>>m=mgrid[1:3,2:5]#rectangularmeshgridwithxvalues[1,2]
andyvalues[2,3,4]
>>>printm
[[[111]

[222]]
[[234]
[234]]]
>>>m[0,1,2]#xvalueofgridpointwithindexcoordinates(1,2)
2
>>>m[1,1,2]#yvalueofgridpointwithindexcoordinates(1,2)
4

Seealso:indices,ogrid

min()
>>>fromnumpyimport*
>>>a=array([10,20,30])
>>>a.min()
10
>>>a=array([[10,50,30],[60,20,40]])
>>>a.min()
10
>>>a.min(axis=0)#foreachofthecolumns,findtheminimum
array([10,20,30])
>>>a.min(axis=1)#foreachoftherows,findtheminimum
array([10,20])
>>>min(a)#alsoexists,butisslower

Seealso:nan,max,minimum,argmin,ptp

minimum()
>>>fromnumpyimport*
>>>a=array([1,0,5])
>>>b=array([3,2,4])
>>>minimum(a,b)#elementbyelementcomparison
array([1,0,4])
>>>min(a.tolist(),b.tolist())#StandardPythonfunctiondoesnot
givethesame!
[1,0,5]

Seealso:min,maximum,argmin

multiply()
>>>fromnumpyimport*
>>>multiply(array([3,6]),array([4,7]))
array([12,42])

Seealso:dot

nan

>>>fromnumpyimport*
>>>sqrt(array([1.0]))
array([nan])#nan=NaN=NotANumber
>>>x=array([2,nan,1])
>>>isnan(x)#showwhichelementsarenan
array([False,True,False],dtype=bool)
>>>isfinite(x)#showwhichelementsarenotnan/inf/inf
array([True,False,True],dtype=bool)
>>>nansum(x)#sameassum()butignorenanelements
3.0
>>>nanmax(x)#sameasmax()butignorenanelements
2.0
>>>nanmin(x)#sameasmin()butignorenanelements
1.0
>>>nanargmin(x)#sameasargmin()butignorenanelements
2
>>>nanargmax(x)#sameasargmax()butignorenanelements
0
>>>nan_to_num(x)#replaceallnanelementswith0.0
array([2.,0.,1.])

Seealso:inf

ndenumerate()
>>>fromnumpyimport*
>>>a=arange(9).reshape(3,3)+10
>>>a
array([[10,11,12],
[13,14,15],
[16,17,18]])
>>>b=ndenumerate(a)
>>>forposition,valueinb:printposition,value#positionis
theNdimensionalindex
...
(0,0)10
(0,1)11
(0,2)12
(1,0)13
(1,1)14
(1,2)15
(2,0)16
(2,1)17
(2,2)18

Seealso:broadcast,ndindex

ndim

>>>fromnumpyimport*
>>>a=arange(12).reshape(3,4)
>>>a
array([[0,1,2,3],
[4,5,6,7],
[8,9,10,11]])
>>>a.ndim#ahas2axes
2
>>>a.shape=(2,2,3)
array([[[0,1,2],
[3,4,5]],
[[6,7,8],
[9,10,11]]])
>>>a.ndim#nowahas3axes
3
>>>len(a.shape)#sameasndim
3

Seealso:shape

ndindex()
>>>forindexinndindex(4,3,2):
printindex
(0,0,0)
(0,0,1)
(0,1,0)
...
(3,1,1)
(3,2,0)
(3,2,1)

Seealso:broadcast,ndenumerate

newaxis
>>>fromnumpyimport*
>>>x=arange(3)
>>>x
array([0,1,2])
>>>x[:,newaxis]#addanewdimension/axis
array([[0],
[1],
[2]])
>>>x[:,newaxis,newaxis]#addtwonewdimensions/axes
array([[[0]],
[[1]],
[[2]]])
>>>x[:,newaxis]*x

array([[0,0,0],
[0,1,2],
[0,2,4]])
>>>y=arange(3,6)
>>>x[:,newaxis]*y#outerproduct,sameasouter(x,y)
array([[0,0,0],
[3,4,5],
[6,8,10]])
>>>x.shape
(3,)
>>>x[newaxis,:].shape#x[newaxis,:]isequivalenttox[newaxis]
andx[None]
(1,3)
>>>x[:,newaxis].shape
(3,1)

Seealso:[],atleast_1d,atleast_2d,atleast_3d,expand_dims

nonzero()
>>>fromnumpyimport*
>>>x=array([1,0,2,1,0,0,8])
>>>indices=x.nonzero()#findtheindicesofthenonzero
elements
>>>indices
(array([0,2,3,6]),)
>>>x[indices]
array([1,2,1,8])
>>>y=array([[0,1,0],[2,0,3]])
>>>indices=y.nonzero()
>>>indices
(array([0,1,1]),array([1,0,2]))
>>>y[indices[0],indices[1]]#onewayofdoingit,explains
what'sinindices[0]andindices[1]
array([1,2,3])
>>>y[indices]#thiswayisshorter
array([1,2,3])
>>>y=array([1,3,5,7])
>>>indices=(y>=5).nonzero()
>>>y[indices]
array([5,7])
>>>nonzero(y)#functionalsoexists
(array([0,1,2,3]),)

Seealso:[],where,compress,choose,take

ogrid()
>>>fromnumpyimport*
>>>x,y=ogrid[0:3,0:3]#xandyareusefultousewith

broadcastingrules
>>>x
array([[0],
[1],
[2]])
>>>y
array([[0,1,2]])
>>>printx*y#examplehowtousebroadcastingrules
[[000]
[012]
[024]]

Seealso:mgrid

ones()
>>>fromnumpyimport*
>>>ones(5)
array([1.,1.,1.,1.,1.])
>>>ones((2,3),int)
array([[1,1,1],
[1,1,1]])

Seealso:ones_like,zeros,empty,eye,identity

ones_like()
>>>fromnumpyimport*
>>>a=array([[1,2,3],[4,5,6]])
>>>ones_like(a)#onesinitialisedarraywiththesameshapeand
datatypeas'a'
array([[1,1,1],
[1,1,1]])

Seealso:ones,zeros_like

outer()
>>>fromnumpyimport*
>>>x=array([1,2,3])
>>>y=array([10,20,30])
>>>outer(x,y)#outerproduct
array([[10,20,30],
[20,40,60],
[30,60,90]])

Seealso:inner,cross

permutation()

>>>fromnumpyimport*
>>>fromnumpy.randomimportpermutation
>>>permutation(4)#permutationofintegersfrom0to3
array([0,3,1,2])
>>>permutation(4)#anotherpermutationofintegersfrom0to3
array([2,1,0,3])
>>>permutation(4)#yetanotherpermutationofintegersfrom0to
3
array([3,0,2,1])

Seealso:shuffle,bytes,seed

piecewise()
>>>fromnumpyimport*
>>>f1=lambdax:x*x
>>>f2=lambdax:2*x
>>>x=arange(2.,3.,0.1)
>>>condition=(x>1)&(x<2)#booleanarray
>>>y=piecewise(x,condition,[f1,1.])#ifconditionistrue,
returnf1,otherwise1.
>>>y=piecewise(x,fabs(x)<=1,[f1,0])+piecewise(x,x>1,
[f2,0])#0.in]inf,1[,f1in[1,+1],f2in]+1,+inf[
>>>printy
<snip>

Seealso:select

pinv()
>>>fromnumpyimport*
>>>fromnumpy.linalgimportpinv,svd,lstsq
>>>A=array([[1.,3.,5.],[2.,4.,6.]])
>>>b=array([1.,3.])
>>>
>>>#Question:findxsuchthat||A*xb||isminimal
>>>#Answer:x=pinvA*b,withpinvAthepseudoinverseofA
>>>
>>>pinvA=pinv(A)
>>>printpinvA
[[1.333333331.08333333]
[0.333333330.33333333]
[0.666666670.41666667]]
>>>x=dot(pinvA,b)
>>>printx
[1.916666670.666666670.58333333]
>>>
>>>#Relationwithleastsquaresminimisationlstsq()

>>>
>>>x,resids,rank,s=lstsq(A,b)
>>>printx#thesamesolutionforxasabove
[1.916666670.666666670.58333333]
>>>
>>>#Relationwithsingularvaluedecompositionsvd()
>>>
>>>U,sigma,V=svd(A)
>>>S=zeros_like(A.transpose())
>>>forninrange(len(sigma)):S[n,n]=1./sigma[n]
>>>dot(V.transpose(),dot(S,U.transpose()))#=pinv(A)
array([[1.33333333,1.08333333],
[0.33333333,0.33333333],
[0.66666667,0.41666667]])

Seealso:inv,lstsq,solve,svd

poisson()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>poisson(lam=0.5,size=(2,3))#poissondistributionlambda=0.5
array([[2,0,0],
[1,1,0]])

Seealso:random_sample,uniform,standard_normal,seed

poly1d()
>>>fromnumpyimport*
>>>p1=poly1d([2,3],r=1)#specifypolynomialbyitsroots
>>>printp1
2
1x5x+6
>>>p2=poly1d([2,3],r=0)#specifypolynomialbyits
coefficients
>>>printp2
2x+3
>>>printp1+p2#+,,*,/andeven**aresupported
2
1x3x+9
>>>quotient,remainder=p1/p2#divisiongivesatupplewiththe
quotientandremainder
>>>printquotient,remainder
0.5x3
15
>>>p3=p1*p2
>>>printp3
32
2x7x3x+18

>>>p3([1,2,3,4])#evaluatethepolynomialinthevalues
[1,2,3,4]
array([10,0,0,22])
>>>p3[2]#thecoefficientofx**2
7
>>>p3.r#therootsofthepolynomial
array([1.5,3.,2.])
>>>p3.c#thecoefficientsofthepolynomial
array([2,7,3,18])
>>>p3.o#theorderofthepolynomial
3
>>>printp3.deriv(m=2)#the2ndderivativeofthepolynomial
12x14
>>>printp3.integ(m=2,k=[1,2])#integratepolynomialtwiceand
use[1,2]asintegrationconstants
5432
0.1x0.5833x0.5x+9x+1x+2

polyfit()
>>>fromnumpyimport*
>>>x=array([1,2,3,4,5])
>>>y=array([6,11,18,27,38])
>>>polyfit(x,y,2)#fita2nddegreepolynomialtothedata,
resultisx**2+2x+3
array([1.,2.,3.])
>>>polyfit(x,y,1)#fita1stdegreepolynomial(straightline),
resultis8x4
array([8.,4.])

Seealso:lstsq

prod()
>>>fromnumpyimport*
>>>a=array([1,2,3])
>>>a.prod()#1*2*3=6
6
>>>prod(a)#alsoexists
6
>>>a=array([[1,2,3],[4,5,6]])
>>>a.prod(dtype=float)#specifytypeofoutput
720.0
>>>a.prod(axis=0)#foreachofthe3columns:product
array([4,10,18])
>>>a.prod(axis=1)#foreachofthetworows:product
array([6,120])

Seealso:cumprod,sum

ptp()
>>>fromnumpyimport*
>>>a=array([5,15,25])
>>>a.ptp()#peaktopeak=maximumminimum
20
>>>a=array([[5,15,25],[3,13,33]])
>>>a.ptp()
30
>>>a.ptp(axis=0)#peaktopeakvalueforeachofthe3columns
array([2,2,8])
>>>a.ptp(axis=1)#peaktopeakvalueforeachofthe2rows
array([20,30])

Seealso:max,min

put()
>>>fromnumpimport*
>>>a=array([10,20,30,40])
>>>a.put([60,70,80],[0,3,2])#firstvalues,thenindices
>>>a
array([60,20,80,70])
>>>a[[0,3,2]]=[60,70,80]#sameeffect
>>>a.put([40,50],[0,3,2,1])#ifvaluearrayistooshort,itis
repeated
>>>a
array([40,50,40,50])
>>>put(a,[0,3],[90])#alsoexists,buthereFIRSTindices,
THENvalues
>>>a
array([90,50,40,90])

Seealso:putmask,take

putmask()
>>>fromnumpimport*
>>>a=array([10,20,30,40])
>>>mask=array([True,False,True,True])#sizemask=sizea
>>>a.putmask([60,70,80,90],mask)#firstvalues,thenthemask
>>>a
array([60,20,80,90])
>>>a=array([10,20,30,40])
>>>a[mask]#reference
array([60,80,90])
>>>a[mask]=array([60,70,80,90])#NOTexactlythesameas
putmask
>>>a

array([60,20,70,80])
>>>a.putmask([10,90],mask)#ifvaluearrayistooshort,itis
repeated
>>>a
array([10,20,10,90])
>>>putmask(a,mask,[60,70,80,90])#alsoexists,buthereFIRST
mask,THENvalues

Seealso:put,take

r_[]
>>>fromnumpyimport*
>>>r_[1:5]#sameasarange(1,5)
array([1,2,3,4])
>>>r_[1:10:4]#sameasarange(1,10,4)
array([1,5,9])
>>>r_[1:10:4j]#sameaslinspace(1,10,4),4equallyspaced
elementsbetween1and10inclusive
array([1.,4.,7.,10.])
>>>r_[1:5,7,1:10:4]#sequencesseparatedwithcommasare
concatenated
array([1,2,3,4,7,1,5,9])
>>>r_['r',1:3]#returnamatrix.If1d,resultisa1xNmatrix
matrix([[1,2]])
>>>r_['c',1:3]#returnamatrix.If1d,resultisaNx1matrix
matrix([[1],
[2]])
>>>a=array([[1,2,3],[4,5,6]])
>>>r_[a,a]#concatenationalong1st(default)axis(rowwise,
that'swhyit'scalledr_)
array([[1,2,3],
[4,5,6],
[1,2,3],
[4,5,6]])
>>>r_['1',a,a]#concatenationalonglastaxis,sameasc_[a,a]
array([[1,2,3,1,2,3],
[4,5,6,4,5,6]])

Seealso:c_,s_,arange,linspace,hstack,vstack,column_stack,concatenate,bmat

rand()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>rand(3,2)
array([[0.65159297,0.78872335],
[0.09385959,0.02834748],
[0.8357651,0.43276707]])

Seealso:random_sample,seed

randint()
Synonymforrandom_integers()
Seerandom_integers

randn()
>>>randn(2,3)
array([[1.22497074,0.29508896,0.75040033],
[0.54822685,0.98032155,1.40467696]])

Seealso:standard_normal,poisson,seed

random_integers()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>random_integers(1,5,(2,2))
array([[3,1],
[1,0]])

Seealso:random_sample,uniform,poisson,seed

random_sample()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>random_sample((3,2))
array([[0.76228008,0.00210605],
[0.44538719,0.72154003],
[0.22876222,0.9452707]])

Seealso:ranf,sample,rand,seed

ranf()
Synonymforrandom_sample
Seerandom_sample,sample

ravel()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4]])
>>>a.ravel()#1dversionofa
array([1,2,3,4])
>>>b=a[:,0].ravel()#a[:,0]doesnotoccupyasinglememory

segment,thusbisacopy,notareference
>>>b
array([1,3])
>>>c=a[0,:].ravel()#a[0,:]occupiesasinglememorysegment,
thuscisareference,notacopy
>>>c
array([1,2])
>>>b[0]=1
>>>c[1]=2
>>>a
array([[1,2],
[3,4]])
>>>ravel(a)#alsoexists

Seealso:flatten

real
>>>fromnumpyimport*
>>>a=array([1+2j,3+4j,5+6j])
>>>a.real
array([1.,3.,5.])
>>>a.real=9
>>>a
array([9.+2.j,9.+4.j,9.+6.j])
>>>a.real=array([9,8,7])
>>>a
array([9.+2.j,8.+4.j,7.+6.j])

Seealso:imag,angle

recarray()
>>>fromnumpyimport*
>>>num=2
>>>a=recarray(num,formats='i4,f8,f8',names='id,x,y')
>>>a['id']=[3,4]
>>>a['id']
array([3,4])
>>>a=rec.fromrecords([(35,1.2,7.3),(85,9.3,3.2)],
names='id,x,y')#fromrecordsisinthenumpy.recsubmodule
>>>a['id']
array([35,85])

Seealso:array,dtype

reduce()
>>>fromnumpyimport*

>>>add.reduce(array([1.,2.,3.,4.]))#computes
((((1.)+2.)+3.)+4.)
10.0
>>>multiply.reduce(array([1.,2.,3.,4.]))#worksalsowithother
operands.Computes((((1.)*2.)*3.)*4.)
24.0
>>>add.reduce(array([[1,2,3],[4,5,6]]),axis=0)#reduceevery
columnseparately
array([5,7,9])
>>>add.reduce(array([[1,2,3],[4,5,6]]),axis=1)#reduceevery
rowseparately
array([6,15])

Seealso:accumulate,sum,prod

repeat()
>>>fromnumpyimport*
>>>repeat(7.,4)
array([7.,7.,7.,7.])
>>>a=array([10,20])
>>>a.repeat([3,2])
array([10,10,10,20,20])
>>>repeat(a,[3,2])#alsoexists
>>>a=array([[10,20],[30,40]])
>>>a.repeat([3,2,1,1])
array([10,10,10,20,20,30,40])
>>>a.repeat([3,2],axis=0)
array([[10,20],
[10,20],
[10,20],
[30,40],
[30,40]])
>>>a.repeat([3,2],axis=1)
array([[10,10,10,20,20],
[30,30,30,40,40]])

Seealso:tile

reshape()
>>>fromnumpyimport*
>>>x=arange(12)
>>>x.reshape(3,4)#arraywith3rowsand4columns.3x4=12.
Totalnumberofelementsisalwaysthesame.
array([[0,1,2,3],
[4,5,6,7],
[8,9,10,11]])
>>>x.reshape(3,2,2)#3x2x2array3x2x2=12.xitselfdoes
_not_change.

array([[[0,1],
[2,3]],
[[4,5],
[6,7]],
[[8,9],
[10,11]]])
>>>x.reshape(2,1)#'missing'1valueniscalculatedsothat
2xn=12,son=6
array([[0,1,2,3,4,5],
[6,7,8,9,10,11]])
>>>x.reshape(12)#reshape(1,12)isnotthesameasreshape(12)
array([0,1,2,3,4,5,6,7,8,9,10,11])
>>>reshape(x,(2,6))#Separatefunctionreshape()alsoexists

Seealso:shape,resize

resize()
>>>fromnumpyimport*
>>>a=array([1,2,3,4])
>>>a.resize(2,2)#changesshapeof'a'itself
>>>printa
[[12]
[34]]
>>>a.resize(3,2)#reallocatesmemoyof'a'tochangenrof
elements,fillsexcesselementswith0
>>>printa
[[12]
[34]
[00]]
>>>a.resize(2,4)
>>>printa
[[1234]
[0000]]
>>>a.resize(2,1)#throwsawayelementsof'a'tofitnewshape
>>>printa
[[1]
[2]]

But,thereisacaveat:
>>>b=array([1,2,3,4])
>>>c=b#cisreferencetob,itdoesn't'own'itsdata
>>>c.resize(2,2)#noproblem,nrofelementsdoesn'tchange
>>>c.resize(2,3)#doesn'twork,cisonlyareference
Traceback(mostrecentcalllast):
File"<stdin>",line1,in?
ValueError:cannotresizeanarraythathasbeenreferencedoris
referencing
anotherarrayinthisway.Usetheresizefunction

>>>b.resize(2,3)#doesn'twork,bisreferencedbyanotherarray
Traceback(mostrecentcalllast):
File"<stdin>",line1,in?
ValueError:cannotresizeanarraythathasbeenreferencedoris
referencing
anotherarrayinthisway.Usetheresizefunction

andit'snotalwaysobviouswhatthereferenceis:
>>>d=arange(4)
>>>d
array([0,1,2,3])
>>>d.resize(5)#doesn'twork,butwhere'sthereference?
Traceback(mostrecentcalllast):
File"<stdin>",line1,in?
ValueError:cannotresizeanarraythathasbeenreferencedoris
referencing
anotherarrayinthisway.Usetheresizefunction
>>>_#'_'wasareferencetod!
array([0,1,2,3])
>>>d=resize(d,5)#thisdoeswork,however
>>>d
array([0,1,2,3,0])

Seealso:reshape

rollaxis()
>>>fromnumpyimport*
>>>a=arange(3*4*5).reshape(3,4,5)
>>>a.shape
(3,4,5)
>>>b=rollaxis(a,1,0)#transposearraysothataxis1is
'rolled'beforeaxis0
>>>b.shape
(4,3,5)
>>>b=rollaxis(a,0,2)#transposearraysothataxis0is
'rolled'beforeaxis2
>>>b.shape
(4,3,5)

Seealso:swapaxes,transpose

round()
round(decimals=0,out=None)>referencetoroundedvalues.
>>>fromnumpyimport*
>>>array([1.2345,1.647]).round()#roundstheitems.Type
remainsfloat64.

array([1.,2.])
>>>array([1,1]).round()#integerarraysstayastheyare
array([1,1])
>>>array([1.2345,1.647]).round(decimals=1)#roundto1decimal
place
array([1.2,1.6])
>>>array([1.2345+2.34j,1.6470.238j]).round()#bothrealand
complexpartsarerounded
array([1.+2.j,2.0.j])
>>>array([0.0,0.5,1.0,1.5,2.0,2.5]).round()#numpyrounds
x.5tonearesteven.
array([0.,0.,1.,2.,2.,2.])
>>>a=zeros(3,dtype=int)
>>>array([1.2345,1.647,3.141]).round(out=a)#differentoutput
arraysmaybespecified
array([1,2,3])
>>>a#andtheoutputiscasttothenewtype
array([1,2,3])
>>>round_(array([1.2345,1.647]))#round_isthefunctional
form.>acopy.
array([1.,2.])
>>>around(array([1.2345,1.647]))#aroundisanaliasof
round_.
array([1.,2.])

Seealso:ceil,floor,fix,astype

rot90()
>>>fromnumpyimport*
>>>a=arange(12).reshape(4,3)
>>>a
array([[0,1,2],
[3,4,5],
[6,7,8],
[9,10,11]])
>>>rot90(a)#'rotate'thematrix90degrees
array([[2,5,8,11],
[1,4,7,10],
[0,3,6,9]])

Seealso:fliplr,flipud

s_[]
>>>fromnumpyimport*
>>>s_[1:5]#easyslicegenerating.Seer_[]examples.
slice(1,5,None)
>>>s_[1:10:4]
slice(1,10,4)

>>>s_[1:10:4j]
slice(1,10,4j)
>>>s_['r',1:3]#toreturnamatrix.If1d,resultisa1xN
matrix
('r',slice(1,3,None))
>>>s_['c',1:3]#toreturnamatrix.If1d,resultisaNx1
matrix
('c',slice(1,3,None))

Seealso:r_,c_,slice,index_exp

sample()
Synonymforrandom_sample
Seealso:random_sample,ranf

savetxt()
>>>fromnumpyimport*
>>>savetxt("myfile.txt",data)#datais2Darray
>>>savetxt("myfile.txt",x)#xis1Darray.1columninfile.
>>>savetxt("myfile.txt",(x,y))#x,yare1Darrays.2rowsin
file.
>>>savetxt("myfile.txt",transpose((x,y)))#x,yare1Darrays.2
columnsinfile.
>>>savetxt("myfile.txt",transpose((x,y)),fmt='%6.3f')#usenew
formatinsteadof'%.18e'
>>>savetxt("myfile.txt",data,delimiter='')#use''to
separatecolumnsinsteadofspace

Seealso:loadtxt,tofile

searchsorted()
searchsorted(keys,side="left")
>>>fromnumpyimport*
>>>a=array([1,2,2,3])#ais1Dandinascendingorder.
>>>a.searchsorted(2)#sidedefaultsto"left"
1#a[1]isthefirstelementina>=2
>>>a.searchsorted(2,side='right')#lookfortheotherendof
therunoftwos
3#a[3]isthefirstelementina>2
>>>a.searchsorted(4)#4isgreaterthananyelementina
4#thereturnedindexis1pasttheendofa.
>>>a.searchsorted([[1,2],[2,3]])#whoa,fancykeys
array([[0,1],#thereturnedarrayhasthesameshapeasthekeys
[1,3]])
>>>searchsorted(a,2)#thereisafunctionalform
1

Seealso:sort,histogram

seed()
>>>seed([1])#seedthepseudorandomnumbergenerator
>>>rand(3)
array([0.13436424,0.84743374,0.76377462])
>>>seed([1])
>>>rand(3)
array([0.13436424,0.84743374,0.76377462])
>>>rand(3)
array([0.25506903,0.49543509,0.44949106])

select()
>>>fromnumpyimport*
>>>x=array([5.,2.,1.,0.,4.,1.,3.,10.])
>>>select([x<0,x==0,x<=5],[x0.1,0.0,x+0.2],default=
100.)
array([5.2,2.1,1.2,0.,4.2,1.1,3.2,100.])
>>>
>>>#Thisishowitworks:
>>>
>>>result=zeros_like(x)
>>>forninrange(len(x)):
...ifx[n]<0:result[n]=x[n]0.1#Theorderofthe
conditionsmatters.Thefirstonethat
...elifx[n]==0:result[n]=0.0#matches,willbe'selected'.
...elifx[n]<=5:result[n]=x[n]+0.2
...else:result[n]=100.#Thedefaultisusedwhennoneofthe
conditionsmatch
...
>>>result
array([5.2,2.1,1.2,0.,4.2,1.1,3.2,100.])

Seealso:choose,piecewise

set_printoptions()
>>>fromnumpyimport*
>>>x=array([pi,1.e200])
>>>x
array([3.14159265e+000,1.00000000e200])
>>>set_printoptions(precision=3,suppress=True)#3digitsbehind
decimalpoint+suppresssmallvalues
>>>x
array([3.142,0.])
>>>

>>>help(set_printoptions)#seehelp()forkeywords
'threshold','edgeitems'and'linewidth'

shape
>>>fromnumpyimport*
>>>x=arange(12)
>>>x.shape
(12,)
>>>x.shape=(3,4)#arraywith3rowsand4columns.3x4=12.
Totalnumberofelementsisalwaysthesame.
>>>x
array([[0,1,2,3],
[4,5,6,7],
[8,9,10,11]])
>>>x.shape=(3,2,2)#3x2x2array3x2x2=12.xitself_does_
change,unlikereshape().
>>>x
array([[[0,1],
[2,3]],
[[4,5],
[6,7]],
[[8,9],
[10,11]]])
>>>x.shape=(2,1)#'missing'1valueniscalculatedsothat
2xn=12,son=6
>>>x
array([[0,1,2,3,4,5],
[6,7,8,9,10,11]])
>>>x.shape=12#x.shape=(1,12)isnotthesameasx.shape=
12
>>>x
array([0,1,2,3,4,5,6,7,8,9,10,11])

Seealso:reshape

shuffle()
>>>fromnumpyimport*
>>>fromnumpy.randomimportshuffle
>>>x=array([1,50,1,3])
>>>shuffle(x)#shuffletheelementsofx
>>>printx
[13501]
>>>x=['a','b','c','z']
>>>shuffle(x)#workswithanysequence
>>>printx
['a','c','z','b']

Seealso:permutation,bytes

slice()
>>>s=slice(3,9,2)#sliceobjectsexistoutsidenumpy
>>>fromnumpyimport*
>>>a=arange(20)
>>>a[s]
array([3,5,7])
>>>a[3:9:2]#samething
array([3,5,7])

Seealso:[],...,newaxis,s_,ix_,indices,index_exp

solve()
>>>fromnumpyimport*
>>>fromnumpy.linalgimportsolve
>>>
>>>#Thesystemofequationswewanttosolvefor(x0,x1,x2):
>>>#3*x0+1*x1+5*x2=6
>>>#1*x0+8*x2=7
>>>#2*x0+1*x1+4*x2=8
>>>
>>>a=array([[3,1,5],[1,0,8],[2,1,4]])
>>>b=array([6,7,8])
>>>x=solve(a,b)
>>>printx#Thisisoursolution
[3.285714299.428571431.28571429]
>>>
>>>dot(a,x)#Justcheckingifweindeedobtaintherighthand
side
array([6.,7.,8.])

Seealso:inv

sometrue()
>>>fromnumpyimport*
>>>b=array([True,False,True,True])
>>>sometrue(b)
True
>>>a=array([1,5,2,7])
>>>sometrue(a>=5)
True

Seealso:alltrue,all,any

sort()

sort(axis=1,kind="quicksort")
>>>fromnumpyimport*
>>>a=array([2,0,8,4,1])
>>>a.sort()#inplacesortingwithquicksort(default)
>>>a
array([0,1,2,4,8])
>>>a.sort(kind='mergesort')#algorithmoptionsare'quicksort',
'mergesort'and'heapsort'
>>>a=array([[8,4,1],[2,0,9]])
>>>a.sort(axis=0)
>>>a
array([[2,0,1],
[8,4,9]])
>>>a=array([[8,4,1],[2,0,9]])
>>>a.sort(axis=1)#defaultaxis=1
>>>a
array([[1,4,8],
[0,2,9]])
>>>sort(a)#thereisafunctionalform

Seealso:argsort,lexsort

split()
>>>fromnumpyimport*
>>>a=array([[1,2,3,4],[5,6,7,8]])
>>>split(a,2,axis=0)#splitain2parts.rowwise
array([[1,2,3,4]]),array([[5,6,7,8]])]
>>>split(a,4,axis=1)#splitain4parts,columnwise
[array([[1],
[5]]),array([[2],
[6]]),array([[3],
[7]]),array([[4],
[8]])]
>>>split(a,3,axis=1)#impossibletosplitin3equalparts>
error(SEE:array_split)
Traceback(mostrecentcalllast):
<snip>
ValueError:arraysplitdoesnotresultinanequaldivision
>>>split(a,[2,3],axis=1)#makeasplitbeforethe2ndandthe
3rdcolumn
[array([[1,2],
[5,6]]),array([[3],
[7]]),array([[4],
[8]])]
ERROR:EOFinmultilinestatement

Seealso:dsplit,hsplit,vsplit,array_split,concatenate

squeeze()
>>>fromnumpyimport*
>>>a=arange(6)
>>>a=a.reshape(1,2,1,1,3,1)
>>>a
array([[[[[[0],
[1],
[2]]]],
[[[[3],
[4],
[5]]]]]])
>>>a.squeeze()#resulthasshape2x3,alldimensionswithlength
1areremoved
array([[0,1,2],
[3,4,5]])
>>>squeeze(a)#alsoexists

std()
>>>fromnumpyimport*
>>>a=array([1.,2,7])
>>>a.std()#normalizedbyN(notN1)
2.6246692913372702
>>>a=array([[1.,2,7],[4,9,6]])
>>>a.std()
2.793842435706702
>>>a.std(axis=0)#standarddeviationofeachofthe3columns
array([1.5,3.5,0.5])
>>>a.std(axis=1)#standarddeviationofeachofthe2columns
array([2.62466929,2.05480467])

Seealso:mean,var,cov

standard_normal()
>>>standard_normal((2,3))
array([[1.12557608,0.13464922,0.35682992],
[1.54090277,1.21551589,1.82854551]])

Seealso:randn,uniform,poisson,seed

sum()
>>>fromnumpyimport*
>>>a=array([1,2,3])

>>>a.sum()
6
>>>sum(a)#alsoexists
>>>a=array([[1,2,3],[4,5,6]])
>>>a.sum()
21
>>>a.sum(dtype=float)#specifytypeofoutput
21.0
>>>a.sum(axis=0)#sumoverrowsforeachofthe3columns
array([5,7,9])
>>>a.sum(axis=1)#sumovercolumnsforeachofthe2rows
array([6,15])

Seealso:accumulate,nan,cumsum,prod

svd()
>>>fromnumpyimport*
>>>fromnumpy.linalgimportsvd
>>>A=array([[1.,3.,5.],[2.,4.,6.]])#Aisa(2x3)matrix
>>>U,sigma,V=svd(A)
>>>printU#Uisa(2x2)unitarymatrix
[[0.619629480.78489445]
[0.784894450.61962948]]
>>>printsigma#nonzerodiagonalelementsofSigma
[9.525518090.51430058]
>>>printV#Visa(3x3)unitarymatrix
[[0.22984770.524744820.81964194]
[0.883461020.240782490.40189603]
[0.408248290.816496580.40824829]]
>>>Sigma=zeros_like(A)#constructingSigmafromsigma
>>>n=min(A.shape)
>>>Sigma[:n,:n]=diag(sigma)
>>>printdot(U,dot(Sigma,V))#A=U*Sigma*V
[[1.3.5.]
[2.4.6.]]

Seealso:pinv

swapaxes()
>>>fromnumpyimport*
>>>a=arange(30)
>>>a=a.reshape(2,3,5)
>>>a
array([[[0,1,2,3,4],
[5,6,7,8,9],
[10,11,12,13,14]],
[[15,16,17,18,19],
[20,21,22,23,24],

[25,26,27,28,29]]])
>>>b=a.swapaxes(1,2)#swapthe2ndandthe3rdaxis
>>>b
array([[[0,5,10],
[1,6,11],
[2,7,12],
[3,8,13],
[4,9,14]],
[[15,20,25],
[16,21,26],
[17,22,27],
[18,23,28],
[19,24,29]]])
>>>b.shape
(2,5,3)
>>>b[0,0,0]=1#beawarethatbisareference,notacopy
>>>printa[0,0,0]

Seealso:transpose,rollaxis

T
>>>fromnumpyimport*
>>>x=array([[1.,2.],[3.,4.]])
>>>x
array([[1.,2.],
[3.,4.]])
>>>x.T#shortcutfortranspose()
array([[1.,3.],
[2.,4.]])

Seealso:transpose

take()
>>>fromnumpyimport*
>>>a=array([10,20,30,40])
>>>a.take([0,0,3])#[0,0,3]isasetofindices
array([10,10,40])
>>>a[[0,0,3]]#thesameeffect
array([10,10,40])
>>>a.take([[0,1],[0,1]])#shapeofreturnarraydependsonshape
ofindicesarray
array([[10,20],
[10,20]])
>>>a=array([[10,20,30],[40,50,60]])
>>>a.take([0,2],axis=1)
array([[10,30],
[40,60]])
>>>take(a,[0,2],axis=1)#alsoexists

Seealso:[],put,putmask,compress,choose

tensordot()
>>>fromnumpyimport*
>>>a=arange(60.).reshape(3,4,5)
>>>b=arange(24.).reshape(4,3,2)
>>>c=tensordot(a,b,axes=([1,0],[0,1]))#sumoverthe1stand
2nddimensions
>>>c.shape
(5,2)
>>>#Aslowerbutequivalentwayofcomputingthesame:
>>>c=zeros((5,2))
>>>foriinrange(5):
...forjinrange(2):
...forkinrange(3):
...forninrange(4):
...c[i,j]+=a[k,n,i]*b[n,k,j]
...

Seealso:dot

tile()
>>>fromnumpyimport*
>>>a=array([10,20])
>>>tile(a,(3,2))#concatenate3x2copiesofatogether
array([[10,20,10,20],
[10,20,10,20],
[10,20,10,20]])
>>>tile(42.0,(3,2))#worksforscalars,too
array([[42.,42.],
[42.,42.],
[42.,42.]])
>>>tile([[1,2],[4,8]],(2,3))#worksfor2darraysandlist
literals,too
array([[1,2,1,2,1,2],
[4,8,4,8,4,8],
[1,2,1,2,1,2],
[4,8,4,8,4,8]])

Seealso:hstack,vstack,r_,c_,concatenate,repeat

tofile()
>>>fromnumpyimport*
>>>x=arange(10.)
>>>y=x**2

>>>y.tofile("myfile.dat")#binaryformat
>>>y.tofile("myfile.txt",sep='',format="%e")#asciiformat,
onerow,expnotation,valuesseparatedby1space
>>>y.tofile("myfile.txt",sep='\n',format="%e")#ascii
format,onecolumn,exponentialnotation

Seealso:fromfile,loadtxt,savetxt

tolist()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4]])
>>>a.tolist()#converttoastandardpythonlist
[[1,2],[3,4]]

trace()
>>>fromnumpyimport*
>>>a=arange(12).reshape(3,4)
>>>a
array([[0,1,2,3],
[4,5,6,7],
[8,9,10,11]])
>>>a.diagonal()
array([0,5,10])
>>>a.trace()
15
>>>a.diagonal(offset=1)
array([1,6,11])
>>>a.trace(offset=1)
18

Seealso:diag,diagonal

transpose()
Averysimpleexample:
>>>a=array([[1,2,3],[4,5,6]])
>>>printa.shape
(2,3)
>>>b=a.transpose()
>>>printb
[[14]
[25]
[36]]
>>>printb.shape
(3,2)

Fromthis,amoreelaborateexamplecanbeunderstood:
>>>a=arange(30)
>>>a=a.reshape(2,3,5)
>>>a
array([[[0,1,2,3,4],
[5,6,7,8,9],
[10,11,12,13,14]],
[[15,16,17,18,19],
[20,21,22,23,24],
[25,26,27,28,29]]])
>>>b=a.transpose()
>>>b
array([[[0,15],
[5,20],
[10,25]],
[[1,16],
[6,21],
[11,26]],
[[2,17],
[7,22],
[12,27]],
[[3,18],
[8,23],
[13,28]],
[[4,19],
[9,24],
[14,29]]])
>>>b.shape
(5,3,2)
>>>b=a.transpose(1,0,2)#Firstaxis1,thenaxis0,thenaxis
2
>>>b
array([[[0,1,2,3,4],
[15,16,17,18,19]],
[[5,6,7,8,9],
[20,21,22,23,24]],
[[10,11,12,13,14],
[25,26,27,28,29]]])
>>>b.shape
(3,2,5)
>>>b=transpose(a,(1,0,2))#Aseparatetranspose()function
alsoexists

Seealso:T,swapaxes,rollaxis

tri()
>>>fromnumpyimport*
>>>tri(3,4,k=0,dtype=float)#3x4matrixofFloats,triangular,

thek=0thdiagonalandbelowis1,theupperpartis0
array([[1.,0.,0.,0.],
[1.,1.,0.,0.],
[1.,1.,1.,0.]])
>>>tri(3,4,k=1,dtype=int)
array([[1,1,0,0],
[1,1,1,0],
[1,1,1,1]])

Seealso:tril,triu

tril()
>>>fromnumpyimport*
>>>a=arange(10,100,10).reshape(3,3)
>>>a
array([[10,20,30],
[40,50,60],
[70,80,90]])
>>>tril(a,k=0)
array([[10,0,0],
[40,50,0],
[70,80,90]])
>>>tril(a,k=1)
array([[10,20,0],
[40,50,60],
[70,80,90]])

Seealso:tri,triu

trim_zeros()
>>>fromnumpyimport*
>>>x=array([0,0,0,1,2,3,0,0])
>>>trim_zeros(x,'f')#removezerosatthefront
array([1,2,3,0,0])
>>>trim_zeros(x,'b')#removezerosattheback
array([0,0,0,1,2,3])
>>>trim_zeros(x,'bf')#removezerosatthebackandthefront
array([1,2,3])

Seealso:compress

triu()
>>>fromnumpyimport*
>>>a=arange(10,100,10).reshape(3,3)
>>>a
array([[10,20,30],

[40,50,60],
[70,80,90]])
>>>triu(a,k=0)
array([[10,20,30],
[0,50,60],
[0,0,90]])
>>>triu(a,k=1)
array([[0,20,30],
[0,0,60],
[0,0,0]])

Seealso:tri,tril

typeDict()
>>>fromnumpyimport*
>>>typeDict['short']
<type'numpy.int16'>
>>>typeDict['uint16']
<type'numpy.uint16'>
>>>typeDict['void']
<type'numpy.void'>
>>>typeDict['S']
<type'numpy.string_'>

Seealso:dtype,cast

uniform()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>uniform(low=0,high=10,size=(2,3))#uniformnumbersinrange
[0,10)
array([[6.66689951,4.50623001,4.69973967],
[6.52977732,3.24688284,5.01917021]])

Seealso:standard_normal,poisson,seed

unique()
>>>fromnumpyimport*
>>>x=array([2,3,2,1,0,3,4,0])
>>>unique(x)#removedoublevalues
array([0,1,2,3,4])

Seealso:compress,unique1d

unique1d()

>>>np.unique1d([1,1,2,2,3,3])
array([1,2,3])
>>>a=np.array([[1,1],[2,3]])
>>>np.unique1d(a)
array([1,2,3])
>>>np.unique1d([1,2,6,4,2,3,2],return_index=True)
(array([1,2,3,4,6]),array([0,1,5,3,2]))
>>>x=[1,2,6,4,2,3,2]
>>>u,i=np.unique1d(x,return_inverse=True)
>>>u
array([1,2,3,4,6])
>>>i
array([0,1,4,3,1,2,1])
>>>[u[p]forpini]
[1,2,6,4,2,3,2]

Seealso:compress,unique

vander()
>>>fromnumpyimport*
>>>x=array([1,2,3,5])
>>>N=3
>>>vander(x,N)#Vandermondematrixofthevectorx
array([[1,1,1],
[4,2,1],
[9,3,1],
[25,5,1]])
>>>column_stack([x**(N1i)foriinrange(N)])#tounderstand
whataVandermondematrixcontains
array([[1,1,1],
[4,2,1],
[9,3,1],
[25,5,1]])

var()
>>>fromnumpyimport*
>>>a=array([1,2,7])
>>>a.var()#normalisedwithN(notN1)
6.8888888888888875
>>>a=array([[1,2,7],[4,9,6]])
>>>a.var()
7.8055555555555571
>>>a.var(axis=0)#thevarianceofeachofthe3columns
array([2.25,12.25,0.25])
>>>a.var(axis=1)#thevarianceofeachofthe2rows
array([6.88888889,4.22222222])

Seealso:cov,std,mean

vdot()
>>>fromnumpyimport*
>>>x=array([1+2j,3+4j])
>>>y=array([5+6j,7+8j])
>>>vdot(x,y)#conj(x)*y=(12j)*(5+6j)+(34j)*(7+8j)
(708j)

Seealso:dot,inner,cross,outer

vectorize()
>>>fromnumpyimport*
>>>defmyfunc(x):
...ifx>=0:returnx**2
...else:returnx
...
>>>myfunc(2.)#worksfine
4.0
>>>myfunc(array([2,2]))#doesn'twork,tryit...
<snip>
>>>vecfunc=vectorize(myfunc,otypes=[float])#declarethe
returntypeasfloat
>>>vecfunc(array([2,2]))#worksfine!
array([2.,4.])

Seealso:apply_along_axis,apply_over_axes

view()
>>>fromnumpyimport*
>>>a=array([1.,2.])
>>>a.view()#newarrayreferringtothesamedataas'a'
array([1.,2.])
>>>a.view(complex)#pretendthataismadeupofcomplexnumbers
array([1.+2.j])
>>>a.view(int)#view(type)isNOTthesameasastype(type)!
array([0,1072693248,0,1073741824])
>>>
>>>mydescr=dtype({'names':['gender','age'],'formats':['S1',
'i2']})
>>>a=array([('M',25),('F',30)],dtype=mydescr)#arraywith
records
>>>b=a.view(recarray)#converttoarecordarray,namesare
nowattributes
>>>>>>a['age']#workswith'a'butnotwith'b'
array([25,30],dtype=int16)

>>>b.age#workswith'b'butnotwith'a'
array([25,30],dtype=int16)

Seealso:copy

vonmises()
>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>vonmises(mu=1,kappa=1,size=(2,3))#VonMisesdistribution
mean=1.0,kappa=1
array([[0.81960554,1.37470839,0.15700173],
[1.2974554,2.94229797,0.32462307]])
>>>frompylabimport*#histogramplotexample
>>>hist(vonmises(1,1,(10000)),50)

Seealso:random_sample,uniform,standard_normal,seed

vsplit()
>>>fromnumpyimport*
>>>a=array([[1,2],[3,4],[5,6],[7,8]])
>>>vsplit(a,2)#split,rowwise,in2equalparts
[array([[1,2],
[3,4]]),array([[5,6],
[7,8]])]
>>>vsplit(a,[1,2])#split,rowwise,beforerow1andbeforerow
2
[array([[1,2]]),array([[3,4]]),array([[5,6],
[7,8]])]

Seealso:split,array_split,dsplit,hsplit,vstack

vstack()
>>>fromnumpyimport*
>>>a=array([1,2])
>>>b=array([[3,4],[5,6]])
>>>vstack((a,b,a))#onlythefirstdimensionofthearraysis
allowedtobedifferent
array([[1,2],
[3,4],
[5,6],
[1,2]])

Seealso:hstack,column_stack,concatenate,dstack,vsplit

weibull()

>>>fromnumpyimport*
>>>fromnumpy.randomimport*
>>>weibull(a=1,size=(2,3))#Ithinkaistheshapeparameter
array([[0.08303065,3.41486412,0.67430149],
[0.41383893,0.93577601,0.45431195]])
>>>frompylabimport*#histogramplotexample
>>>hist(weibull(5,(1000)),50)

Seealso:random_sample,uniform,standard_normal,seed

where()
>>>fromnumpyimport*
>>>a=array([3,5,7,9])
>>>b=array([10,20,30,40])
>>>c=array([2,4,6,8])
>>>where(a<=6,b,c)
array([10,20,6,8])
>>>where(a<=6,b,1)
array([10,20,1,1])
>>>indices=where(a<=6)#returnsatuplethearraycontains
indices.
>>>indices
(array([0,1]),)
>>>b[indices]
array([10,20])
>>>b[a<=6]#analternativesyntax
array([10,20])
>>>d=array([[3,5,7,9],[2,4,6,8]])
>>>where(d<=6)#tuplewithfirstalltherowindices,thenall
thecolumnindices
(array([0,0,1,1,1]),array([0,1,0,1,2]))

Beawareofthedifferencebetweenx[listofbools]andx[listofintegers]!
>>>fromnumpyimport*
>>>x=arange(5,0,1)
>>>printx
[54321]
>>>criterion=(x<=2)|(x>=5)
>>>criterion
array([True,False,False,True,True],dtype=bool)
>>>indices=where(criterion,1,0)
>>>printindices
[10011]
>>>x[indices]#integers!
array([4,5,5,4,4])
>>>x[criterion]#bools!
array([5,2,1])
>>>indices=where(criterion)

>>>printindices
(array([0,3,4]),)
>>>x[indices]
array([5,2,1])

Seealso:[],nonzero,clip

zeros()
>>>fromnumpyimport*
>>>zeros(5)
array([0.,0.,0.,0.,0.])
>>>zeros((2,3),int)
array([[0,0,0],
[0,0,0]])

Seealso:zeros_like,ones,empty,eye,identity

zeros_like()
>>>fromnumpyimport*
>>>a=array([[1,2,3],[4,5,6]])
>>>zeros_like(a)#withzerosinitialisedarraywiththesame
shapeanddatatypeas'a'
array([[0,0,0],
[0,0,0]])

Seealso:ones_like,zeros
CategoryCategory
NumpyExampleList(ltimaedicin2011120820:40:16efectuadapor
SimonBull)