PA196: Pattern Recognition
Exercises 01
Dr. Vlad Popovici
popovici@recetox.muni.cz
RECETOX
Masaryk University, Brno
Python environment
• the easiest is to use a "distribution" of Python
• suggestion/recommendation: Anaconda
http://www.continuum.io/
• Python 2 vs Python 3: for the exercises in the course,
there should be (almost) no difference
• starting with a minimal distribution: use Miniconda:
http://conda.pydata.org/miniconda.html: pick
the right installer (e.g. 64 bit)
Example:
• download the "Python 3.4/64 bit" version of Miniconda
• it may require chmod +x
Miniconda3-3.6.0-Linux-x86_64.sh
• run the installer
• even though the base distribution is Python 3.4 you
can still have Python 2.7 installed as well
• browse through the documentation
• with conda install anaconda you will get all the
packages as in the basic "Anaconda"; alternatively
you can install them on a as-needed basis
• install the machine learning kit: conda install
scikit-learn
About Python language
• tons of source of information
• quick introduction:
https://docs.python.org/2/tutorial/
• more detailed - but free book:
http://www.diveintopython.net/
IPython
• great tool for interactive sessions with Python
• www.ipython.org
• you can have mixed code, text and results in the
same notebook, like in Mathematica
• try: ipython qtconsole -matplotlib inline
and in the console:
from matplotlib import pylab as plt
plt.plot([1,2,3],[4,5,6])
• try (from the command line): ipython notebook
and then open the web page at
http://127.0.0.1:8888
Several key packages:
• numpy: fast array operations and matrix manipulation
• scipy: loads of numerical methods, including some
functions for signal and image processing
• matplotlib: Matlab-style plotting functions - and not
only!
• pandas: versatile package for data analysis
• scikit-image: image processing (beyond scipy)
• scikit-learn: our main interest
Scikit-Learn
• http://scikit-learn.org/
• nice Python package for machine learning/pattern
recognition
• good documentation
• still under development
• start Python: ipython -matplotlib qt
• let’s go through the tutorial at
http://scikit-learn.org/stable/tutorial/
basic/tutorial.html