用Python做科学计算 – Python for scientific computing – Why Python?

View Github Repository Open presentation in a new window

kkhust

See all presentation from kkhust

用Python做科学计算 – Python for scientific computing – Why Python?

1 0

Python-for-scientific-computing

tutorial

On Github kkhust / Python-for-scientific-computing

用Python做科学计算

Python for scientific computing

Created by Yuxiang Chen

What did people say about Python?

“Python is the most powerful language you can still read.”

—Paul Dubois, MySQL

“Python has been an important part of Google since the beginning, and remains so as the system grows and evolves. Today dozens of Google engineers use Python, and we're looking for more people with skills in this language.”

—Peter Norvig, Google

Why Python?

  • 易学
  • 简便
  • 万能
  • Free!

易学

简便

print "hello world"
def add(a, b): # define a function
  return a+b
						
with open('workfile', 'r') as f: # open a file
  line = f.read()
						
class Human: # define a class
  """Human docstring"""
  def walk(self):
    print 'walking ...'
						
x, y = y, x # switching variables
						
for a in 'string': # loop
  print a

万能

What is this talk about?

  • 常用工具 / 库
  • 做图
  • 如何扩展Python
  • 并行计算
  • 其他应用

Things to know first - IDE

  • Eclipse + PyDev
  • Visual Studio (PTVS)
  • PyCharm

建议预装软件包

  • Python(x, y)
  • Enthought Python Distribution (EPD)

必备工具 - IPython

  • 互动
    • Auto-completion
    • Easy access documentation
    • Syntax highlighting
    • History retriever
  • 分享

常用库 - Demo时间！

  • Numpy / Scipy (Matlab for Python)
  • Matplotlib (Scientific plotting)
  • Sympy (Symbolic mathematics)
  • Scikit-learn (Machine learning)
  • Scikit-image (Image processing)
  • Mayavi (3D plotting)

如何写出高效的Python程序

“Premature optimization is the root of all evil.”

—Donald Knuth, TAOCP
  • Profiling!
  • 避免大规模For循环
  • PyPy
  • 扩展Python
  • 并行

Profiling

$  python -m cProfile -o demo.prof demo.py

Use gprof2dot to visualize the profiling

$  gprof2dot -f pstats demo.prof | dot -Tpng -o demo-prof.png

如何扩展Python

  • Python C API / Ctypes
  • Ctypes
  • SWIG
%module swig_math
%{
#define SWIG_FILE_WITH_INIT

int add(double *f, int dim1, double *g, int dim2, double *out, int dim3);

%}
%include "numpy.i"
%init %{
    import_array();
%}

int add(double *IN_ARRAY1, int DIM1, double *IN_ARRAY1, int DIM1, double *INPLACE_ARRAY1, int DIM1);

并行计算

  • 多线程
  • GPU computing - PyCuda
  • Message Passing Interface (MPI)
from mpi import MPI
mpi = MPI()

def pow(x):
    return x**2

mpi.begin()

x = range(10)

result = mpi.parfor(pow, x)

mpi.end()

其他应用

Web框架

  • Django
  • Bottle
  • CherryPy
  • Flask
  • Tornado

其他应用

云计算

只要有你，我就无所不能！

THE END

BY Yuxiang Chen