Resources

I M. Scott Shell,
http://www.engr.ucsb.edu/%7eshell/che210d/numpy.pdf
I http://wiki.scipy.org/NumPy_for_Matlab_Users
I
http://mathesaurus.sourceforge.net/matlab-numpy.html

1 / 12
Numpy arrays

I import numpy as np
I Numpy provides class ndarray, called “array”
I Create array from a list
>>> x = np.array([3.0,5,7,5])
>>> x
array([ 3., 5., 7., 5.])
I If appear to be integers in list, need “float”
I 2D arrays
>>> A = np.array([[8.,1.,6.],[3.,5.,7.],[4.,9.,2.]])
array([[ 8., 1., 6.],
[ 3., 5., 7.],
[ 4., 9., 2.]])

2 / 12
Subscripts
I Use brackets to denote
subscripts
I Start counting at 0
>>> x[0]
3.0
>>> A[1,2]
7.0
>>> x
I Colons work, be careful of last array([ 3., 5., 7., 5.])
value! >>> A
>>> x[0:1] array([[ 8., 1., 6.],
array([ 3.0]) [ 3., 5., 7.],
[ 4., 9., 2.]])

3 / 12
Subscripts
I Use brackets to denote
subscripts
I Start counting at 0
>>> x[0]
3.0
>>> A[1,2]
7.0
>>> x
I Colons work, be careful of last array([ 3., 5., 7., 5.])
value! >>> A
>>> x[0:1] array([[ 8., 1., 6.],
array([ 3.0]) [ 3., 5., 7.],
>>> x[0:2] [ 4., 9., 2.]])
array([ 3.0, 5.])

3 / 12
Subscripts
I Use brackets to denote
subscripts
I Start counting at 0
>>> x[0]
3.0
>>> A[1,2]
7.0
>>> x
I Colons work, be careful of last array([ 3., 5., 7., 5.])
value! >>> A
>>> x[0:1] array([[ 8., 1., 6.],
array([ 3.0]) [ 3., 5., 7.],
>>> x[0:2] [ 4., 9., 2.]])
array([ 3.0, 5.])
>>> A[:,2]
array([ 6., 7., 2.])

3 / 12
Subscripts
I Use brackets to denote
subscripts
I Start counting at 0
>>> x[0]
3.0
>>> A[1,2]
7.0
>>> x
I Colons work, be careful of last array([ 3., 5., 7., 5.])
value! >>> A
>>> x[0:1] array([[ 8., 1., 6.],
array([ 3.0]) [ 3., 5., 7.],
>>> x[0:2] [ 4., 9., 2.]])
array([ 3.0, 5.])
>>> A[:,2]
array([ 6., 7., 2.])
I Negative indices count from end
>>> x[-1]
5.0
3 / 12
Attributes

>>> A.shape
(3, 3)
>>> A.flatten()
array([ 8., 1., 6., 3., 5., 7., 4., 9., 2.])
>>> B=A.copy()
>>> B[1,1]=-1
>>> A[1,1]
5.0
>>> B[1,1]
-1.0
>>> A.transpose()
array([[ 8., 3., 4.],
[ 1., 5., 9.],
[ 6., 7., 2.]])

4 / 12
Methods
>>> x=np.arange(24) # array-range
>>> y=x.reshape([4,6]).copy() # turn into 4 X 6 matrix
>>> y
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]])
>>> np.sum(y) # sum all of y
276
>>> y.sum() # sum all of y
276
>>> y.sum(0) # sum columns
array([36, 40, 44, 48, 52, 56])
>>> y.sum(axis=0) # sum columns
array([36, 40, 44, 48, 52, 56])
>>> np.sum(y,axis=0) # sum columns
array([36, 40, 44, 48, 52, 56])
>>> np.sum(y[:,0]) # sum only first column
36
>>> np.sum(y[:,1]) # sum only second column
40
>>> y.sum(1) # sum along rows
array([ 15, 51, 87, 123])
>>> y.sum(axis=1) # sum along rows
array([ 15, 51, 87, 123])
>>> np.sum(y[0,:]) # sum first row
15

5 / 12
Useful attributes

I π =np.pi, e =np.e
I 1d array of zeros x=np.zeros(N)
I 2d array of zeros x=np.zeros([N,M])
I Array of zeros same size as: y=np.zeros_like(x)
I np.ones, np.ones_like
I np.empty, np.empty_like

6 / 12
Functions like M ATLAB

I np.diag
I np.random.rand
I np.tril, np.triu

7 / 12
Operations on arrays

Suppose x is the array [xi ], y is the array [yi ], A is the 2d array [aij ]
and B is the 2d array [bij ]:
I All operations are elementwise.
I np.sin(x) is the array [sin(xi )], same for matrices
I 3*x is the array [3xi ], same for matrices
I x+y is the array [xi + yi ], same for matrices
I x*y is the array [xi yi ], same for matrices
I x/y is the array [xi /yi ], same for matrices
P
I np.dot(x,y)= i xi yi (usual dot product)
P
I np.dot(A,B)=[ k aik bkj ] (usual matrix multiplication)

8 / 12
Array selection

It is sometimes convenient to select particular elements from arrays.

9 / 12
Array selection by subscript
I One-dimensional arrays
>>> x=2*np.arange(9.)
>>> x
array([ 0., 2., 4., 6., 8., 10., 12., 14., 16.])
>>> i=np.array([0,0,1,5,2])
>>> i
array([0, 0, 1, 5, 2])
>>> x[i]
array([ 0., 0., 2., 10., 4.])
I Two-dimensional arrays: selector for each axis.
>>> A=x.reshape([3,3])
>>> A
array([[ 0. , 2., 4.],
[ 6., 8., 10. ],
[ 12. , 14., 16.]])
# construct minor associated with A[1,0]
>>> i=[[0,0],[2,2]] # 2-axis means result will be 2-axis
>>> j=[[1,2],[1,2]]
>>> A[i,j]
array([[ 2., 4.],
[ 14., 16.]])
10 / 12
Higher-level linear algebra

I import scipy.linalg as la
I Built on BLAS and Lapack
I la.norm: vector and matrix norm
I la.det: determinant
I la.solve: solve system of equations
I la.inv: construct inverse matrix
I la.eig: eigenvalues and eigenvectors
I la.eigvals: eigenvalues only
I fenics does not use this package!

11 / 12
Plotting
I import matplotlib.pyplot as plt
I Must use plt.show() to see your plot!
I Plot like M ATLAB
>>> x=np.linspace(0,4*np.pi,1000)
>>> plt.plot(x,np.exp(-x),x,np.sin(x))
[<matplotlib.lines.Line2D object at 0x4306a90>, <matp
>>> plt.legend((’exp’,’sin’))
<matplotlib.legend.Legend object at 0x45df890>
>>> plt.show()
I plt.semilogy also works

12 / 12