Overview

Numpy
Matplotlib
Scipy

Introduction to Scientific Computing in Python

Numpy, Scipy and Matplotlib packages

Oscar Dalmau

Centro de Investigación en Matemáticas

CIMAT A.C. Mexico

December 2013

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Outline

1 Overview

2 Numpy

3 Matplotlib

4 Scipy
Examples

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Overview

Numpy: provides powerful numerical arrays objects, and

routines to manipulate them
Scipy: high-level data processing routines. Optimization,
regression, interpolation, etc
Matplotlib: 2-D visualization
Mayavi: 3-D visualization

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Creating arrays

Manual construction of array 1D:

>>> import numpy as np
>>> a = np.array([1.,21.,3.])
>>> a.ndim
>>> a.shape
>>> a.size
>>> a.max()
>>> a.argmax()

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Creating arrays

Manual construction of array 2D:

>>> import numpy as np
>>> a = np.array([[1.,21.,3.], [2,44,6]])
>>> a.ndim
>>> a.shape
>>> a.size
>>> a.max()
>>> a.argmax()

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Creating arrays

Manual construction of array 3D:

>>> import numpy as np
>>> a = np.array([[[1.,21.,3.], [2,44,6]],
[[1.,2.,1.], [20,4,2]]])
>>> a.ndim
>>> a.shape
>>> a.size
>>> a.max()
>>> a.argmax()

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Creating arrays

Functions for creating arrays:

>>>a=np.arange(10) # 0 .. n-1
>>>b=np.arange(1,9,2)#start,end(exclusive),step
>>>c=np.linspace(0,1,6)#start,end,num-points
>>>d=np.linspace(0,1,5,endpoint=False)
>>>e=np.ones((3,3))#(3, 3) is a tuple
>>>f=np.zeros((3,3))
>>>g=np.eye(3)
>>>h=np.diag(np.array([1,2,3,4]))
>>>i=np.random.rand(4,3)#uniform in [0, 1]
>>>j=np.random.randn(2,5)#Gaussian
>>>np.random.seed(1234)#Setting the random seed

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Basic data types

bool: bool, bool8, bool

Integers: byte, short, int8, int16, int 32, int64, ...
Unsigned integers: ubyte, ushort, uint8, uint16, uint 32,
uint64, ...
Floating: single, double, float16, float32, float64, float96,
float128
Complex Floating: csingle, complex64, complex128,
complex192, complex256

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Basic data types

>>> import numpy as np

>>> a = np.array([[1.,21.,3.], [2,44,6]],
dtype=int8)
>>> b = np.array([[1.,21.,3.], [2,44,6]],
uint64)

>>> a.ndim
>>> a.shape
>>> a.dtype
>>> b.dtype

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Indexing, slicing and assignment 1D

The items of an array can be accessed and assigned to the

same way as other Python sequences (e.g. lists):
One dimensional arrays:
>>>a=np.arange(10)
>>>a[0]
>>>a[-1] # a[len(a)-1]
>>>a[::2]
>>>a[::-1]
>>>a[0]=10
>>>a[0:2]=[12,16]

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

deleting and inserting 1D

The items of an array can be accessed and assigned to the

same way as other Python sequences (e.g. lists):
One dimensional arrays:
>>>a[0]=[]#Error!!!
>>>a=np.delete(a,0)#delete the first element
>>>a=np.delete(a,-1)#delete the last element
>>>a=np.insert(a,0,5)#array,index,value
# insert the ’value’ before the ’index’

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Indexing and slicing ND

The items of an array can be accessed and assigned to the

same way as other Python sequences (e.g. lists):
Multi dimensional arrays:
>>>a=(10*np.random.rand(5,5)).astype(int)
>>>a[0,0]
>>>a[-1,-1]
>>>a[:,1]
>>>a[::2,::2]
>>>a[0:2,1:4]
>>>a[1:4,3:5]#a[1:4,:][:,3:5]
>>>a[np.ix_([1,3,4],[0,2])]

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Assignment ND

The items of an array can be accessed and assigned to the

same way as other Python sequences (e.g. lists):
Multi dimensional arrays:
>>>a[0,0]*=10
>>>a[1:4,3:5]=100*np.ones((3,2))#a[1:4,3:5]=100
>>>a[np.ix_([1,3],[0,4])]=np.random.rand((3,2))
>>>(a>4).choose(a,4)
>>>a.clip(min=None, max=4)
>>>a.clip(min=3, max=7)
>>>a[:]=3

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Numpy Functions
Deleting and inserting in ND
Multi dimensional arrays:
>>>np.delete(a,0,0)#arr,obj(int),axis
>>>np.delete(a,np.s_[::2],0)#arr,obj(slice),axis
>>>np.delete(a,np.arange(0,4,2),1)#obj(array)
>>>np.delete(a,[1,3])#np.delete(a,[1,3],None)
>>>np.delete(a,[1,3],0)
>>>np.insert(a,1,100,1)#arr,obj(int,slice,seq),
#value,axis
>>>np.insert(a,(1,4),100,1)
>>>np.insert(a,1,[1,2,3,4,5],1)
>>>ia=np.zeros((5,2))
>>>np.insert(a,(1,4),ia,1)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Numpy Functions

Multi dimensional arrays:

>>>a.take([0,2,2], axis=0).take([2,4], axis=1)
>>>a.diagonal(offset=0)#a.diagonal(0)
>>>a.sum(axis=0)#a.sum(0)
>>>a.sum(axis=1)#a.sum(1)
>>>a.sum()
>>>a.trace(offset=0)#a.trace(0)
>>>a.max()# a.max(0), a.max(1)
>>>a.T #a.transpose()
Statistical functions: mean, median, std, var, correlate, cov

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Numpy Functions

Convert ND array into a vector (column or row):

>>>c=a.flatten()#convert into an 1D array
>>>c=a.flatten()[:,np.newaxis]#column vector
>>>c=np.array([2,3,4,5]).reshape(-1,1)#column ve
>>>c=np.r_[’c’,1:10]#column vector,(matrix obj)
>>>(c.T*c)[0,0]#inner product
>>>c*c.T#outer product

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Numpy Functions

Concatenation
>>>a=np.array([[0, 1, 2], [3, 4, 5]])
>>>c=np.concatenate((a,a),0)#np.vstack((a,a))
>>>c=np.concatenate((a,a),1)#np.hstack((a,a))
>>>c=np.r_[’0’, a,a]
>>>c=np.r_[’1’, a,a]

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Vector and matrix mathematics

NumPy provides many functions for performing standard

vector and matrix multiplication routines
>>>a=np.array([1, 2, 3], float)
>>>b=np.array([0, 1, 1], float)
>>>np.dot(a, b)
>>>np.inner(a,b)
>>>np.outer(a,b)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Vector and matrix mathematics

Linear Algebra
>>>a=np.array([[4, 2, 0], [9, 3, 7], [1, 2, 1]],
>>>b=np.linalg.inv(a)
>>>np.dot(a,b)#np.matrix(a)*np.matrix(b)
>>>U,D,V=np.linalg.svd(a)
>>>np.dot(np.dot(U,np.diag(D)),V)#===a
>>>np.matrix(U)*np.matrix(np.diag(D))*\
np.matrix(V)#===a

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Commonly-used numpy.linalg functions

Function Description
diag Return the diagonal (or off-diagonal) elements
dot Matrix multiplication
trace Compute the sum of the diagonal elements
det Compute the matrix determinant
eig Compute the eigenvalues and eigenvectors of
a square matrix
inv Compute the inverse of a square matrix
pinv Compute the Moore-Penrose pseudo-inverse
qr Compute the QR decomposition
svd Compute the singular value decomposition
solve Solve the linear system Ax = b for x
lstsq Compute the least-squares solution to y = Xb,
See exampleleastsquare.py
Oscar Dalmau Image Processing
 Overview
Numpy
Matplotlib
Scipy

Boolean or ‘mask’ index arrays

Boolean index array (result 1D array)

>>>y = np.arange(35).reshape(5,7)
>>>y[y>10] # the result is an ndarray of 1D
Boolean index array (result 2D array)
>>>y = np.arange(35).reshape(5,7)
>>>row1=np.array(np.arange(5))
>>>y[row1>2,:] # select rows with index > 2
>>>row2=np.array([’Bob’,’Joe’,’Will’,
’Bob’,’Will’])
>>>y[row2==’Will’,:]

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Universal Functions: Fast Element-wise Array

Functions

A universal function, or ufunc, is a function that performs

elementwise operations on data in ndarrays. You can think
of them as fast vectorized wrappers for simple functions
that take one or more scalar values and produce one or
more scalar results.
Examples of universal function: abs, fabs, sqrt, square,
exp, log, log10, log2, sign, ceil, floor, modf, isnan, isfinite,
isinf, cos, cosh, sin, sinh, tan, tanh, arccos, arccosh,
arcsin, arcsinh, arctan, arctanh
>>>a = np.array([1,2,3])
>>>np.sqrt(a)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Matplolib

Matplotlib is 2D desktop plotting package. The project was

started by John Hunter in 2002 to enable a MATLAB-like
plotting interface in Python.
Now, matplotlib has a number of add-on toolkits, such as
mplot3d for 3D plots (for me too slow!!!)
Matplotlib is (very) well documented, see
http://matplotlib.org/index.html

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example

>>>import numpy as np
>>>import matplotlib.pyplot as plt
>>>x=np.linspace(-1,1,100);y=x**2
>>>plt.plot(x,y)
>>>plt.show(False)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab

>>>from pylab import *

>>>x=linspace(-1,1,100);y=x**2
>>>plot(x,y)
>>>show(False)
Or using IPython
$ ipython --pylab
In [1]: x=linspace(-1,1,100);y=x**2
In [2]: plot(x,y)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab and IPython

Adding graphics in the same figure

x=linspace(-pi,pi,100);
ys,yc,y2=sin(x),cos(x),x**2
plot(x,ys)
plot(x,yc)
plot(x,y2)
Or simply
x=linspace(-pi, pi, 100)
ys,yc,y2 = sin(x),cos(x),x**2
plot(x,ys,x,yc,x,y2)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab and IPython

Adding style
#Create a new figure of size 8x6 inches, using 100 d
figure(1,figsize=(8,6), dpi=100)#num,figsize,dpi
x=linspace(-pi,pi,100);
s,c=sin(x),cos(x)
plot(x,s,color="blue",linewidth=1.0,linestyle="-")
plot(x,c,color="green",linewidth=1.0,linestyle="-")
#Set x, y limits
xlim(-4.0,4.0); ylim(-1.0,1.0)
#Set x, y ticks
xticks(linspace(-4,4,9)),yticks(linspace(-1,1,5))
#Save figure using 80 dots per inch
savefig("exercice1.png",dpi=80)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab and IPython

MATLAB-style
figure(1,figsize=(8,6), dpi=100)#num,figsize,dpi
x=linspace(-pi,pi,100);
s,c=sin(x),cos(x)
plot(x,s,’b-’,x,c,’g-’)
#Set x, y limits
xlim(-4.0,4.0); ylim(-1.0,1.0)
#Set x, y ticks
xticks(linspace(-4,4,9)),yticks(linspace(-1,1,5))
#Save figure using 80 dots per inch
savefig("exercice1.png",dpi=80)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab and IPython

Adding information: title, labels, legend

title(’Cos and Sin Functions’)
xlabel(’Eje X’)
xlabel(’Eje Y’)
grid()
legend(loc=’upper left’)
xticks( [-pi,-pi/2,0,pi/2,pi])
yticks([-1, 0, +1])
xticks([-pi,-pi/2, 0,pi/2,pi],
[r’$-\pi$’,r’$-\pi/2$’,r’$0$’,r’$+\pi/2$’,r’$

yticks([-1,0,+1],
[r’$-1$’,r’$0$’,r’$+1$’])

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

First Example. Using pylab and IPython

Adding annotations
t = 2*pi/3
plot([t,t],[0,cos(t)],color =’blue’,linewidth=2.5,li
scatter([t,],[cos(t),],50,color=’blue’)
annotate(r’$\sin(\frac{2\pi}{3})=\frac{\sqrt{3}}{2}$
xy=(t,sin(t)),xycoords=’data’,
xytext=(+10,+30),textcoords=’offset points’
arrowprops=dict(arrowstyle="->",connections
plot([t,t],[0,sin(t)],color=’red’,linewidth=2.5,line
scatter([t,],[sin(t),],50,color=’red’)
annotate(r’$\cos(\frac{2\pi}{3})=-\frac{1}{2}$’,
xy=(t,cos(t)),xycoords=’data’,
xytext=(-90,-50),textcoords=’offset points’
arrowprops=dict(arrowstyle="->",connections
Oscar Dalmau Image Processing
 Overview
Numpy
Matplotlib
Scipy

Subplot

Subplots allow us to arrange plots in a regular grid. One

needs to specify the number of rows and columns and the
number of the plot.
Axes are very similar to subplots but allow placement of
plots at any location in the figure. So if we want to put a
smaller plot inside a bigger one we do so with axes.

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Example Subplot

figure(1,figsize=(8,6), dpi=100)#num,figsize,dpi
x=linspace(-pi,pi,100);
s,c,t,y2=sin(x),cos(x),tan(x),x**2
subplot(2,2,1);plot(x,s,’r-’); title(’Sin’)
subplot(2,2,2);plot(x,c,’b-’); title(’Cos’)
subplot(2,2,3);plot(x,t,’g-’); title(’Tan’)
subplot(2,2,4);plot(x,y2,’k-’);title(’Parabola’)

Oscar Dalmau Image Processing

 Overview
Numpy
Matplotlib
Scipy

Some Pyplot functions

figure xlim plot stem

savefig xticks loglog step
subplot ylabel semilogx pie
gca ylim semilogy contour
gcf yscale hist contourf
grid yticks hist2d quiver
title annotate bar imshow
axes arrow barbs spy
axis Circle barh pcolor
legend Polygon boxplot cm
xlabel Rectangle scatter ginput

Oscar Dalmau Image Processing

 Overview
Numpy
Examples
Matplotlib
Scipy

Scipy

scipy package contains the following subpackages:

optimize: routines for optimization
misc: Various utilities that dont have another home (e.g.
lena, imfilter, imresize, imrotate, imread, imshow, imsave,
etc).
ndimage: Multidimensional image processing (e.g., filters,
interpolation, measurements, io).
signal: Signal processing (N-dimensional arrays, e.g.
convolution, correlation, filtering, etc.).
fftpack: Discrete Fourier transforms.
...

Oscar Dalmau Image Processing