Showing posts with label statistics. Show all posts
Showing posts with label statistics. Show all posts

Friday, September 14, 2012

Switching From Perl to Python, Step 5 First Step into Machine Learning

in this post i share my experience while searching about the python machine learning modules. There are lots of them, i think that there is no one that can be used for all algorithms, so for a specific algorithm you can choose one of them that satisfies your need.

First i start reading  Scientific Scripting with Python for Computational Immunology, this is the best, short tutorial ever to understand the basic statistics. While going through stackoverflow questions, i realized that many people recommend scikit-learn: machine learning in Python.

i'm familiar with matplotlib and pyplot, however now in examples another module pylab is imported. Clarification: matplotlib, pyplot, and pylab from (http://truongnghiem.wordpress.com):

pyplot is just a wrapper module to provide a Matlab-style interface to matplotlib.
Many plotting functions in Matlab are provided by pyplot with the same names and arguments.
This will ease the process of moving from Matlab to Python for scientific computation.
pylab is basically a mode in which pyplot and numpy are imported in a single namespace,
thus making the Python working environment very similar to Matlab. By importing pylab: 
from pylab import *
we can use Matlab-style commands like:
x = arange(0, 10, 0.2)
y = sin(x)
plot(x, y)

Ok. Let's start, first i try simple linear regression from Scientific Scripting with Python for Computational Immunology. We have dilution.cvs file that contains the data of:
Dilution Factor,Rep 1,Rep 2,Rep 3,Mean,sd
1,15.16,14.95,14.55,14.89,0.31
2,15.36,15.61,15.51,15.49,0.13
4,16.65,16.88,16.71,16.75,0.12
8,18.07,17.60,18.13,17.93,0.29
16,18.86,19.63,19.39,19.29,0.39
32,20.39,19.40,20.39,20.06,0.57
64,21.44,20.76,21.22,21.14,0.35
128,21.90,22.04,21.94,21.96,0.07
256,22.87,22.77,23.36,23.00,0.32
512,23.98,23.92,24.24,24.05,0.17
1024,24.91,24.83,24.92,24.89,0.05
2048,26.37,25.43,26.21,26.00,0.50
Dilution and Factor columns are used in order to implement the linear regression in two dimensional space where the line is defined as :

'''
 *@author beck 
 *@date Sep 14, 2012
 *Basic Statistics with Python 
 *bekoc.blogspot.com 
'''

import numpy
import matplotlib.pyplot as plt
#import pylab, # from pylab import *
import scipy.stats as stats

xs= numpy.loadtxt("dilution.csv",delimiter=",", skiprows=1, usecols=(0,1)) 
# numpy array - similar to C array notation.
x= numpy.log2(xs[:,0])
y=xs[:,1]

plt.plot(x,y,"x")
        
plt.xlabel('Number of Dilutions(log2)')
plt.ylabel('Rep1')
plt.title('Linear Regression example')
plt.legend()

slope, intercept, r_value, p_value, std_error = stats.linregress(x,y)
plt.plot(x,intercept+slope*x,"r-") # y=mx+b where m is slope and b is intercept
#plt.plot(x,x**2)
plt.show()

straight line seems reasonable

Resources for machine learning:

Tuesday, September 4, 2012

Switching From Perl to Python, Step 4 Basic Statistics


On August 28 2012 at 10am, the creator of matplotlib, John D. Hunter died from complications arising from cancer treatment, after a brief but intense battle with this terrible illness. John is survived by his wife Miriam, his three daughters Rahel, Ava and Clara, his sisters Layne and Mary, and his mother Sarah.
If you have benefited from John's many contributions, please say thanks in the way that would matter most to him. Please consider making a donation to the John Hunter Memorial Fund.

Rest in peace, John  Hunter (1968-August 28th, 2012). My sincere condolences to his family. Thank you for all your contribution. A great loss to the community.Thank you again.  

After learning the basics of Python, today i will try to implement some basic statistical methods including plotting. We need two important packages mathplotlib and scipy. I remember that there is also numpy library, a bit confused but it's explained very well in official scipy documentation:
"SciPy (pronounced "Sigh Pie") is open-source software for mathematics, science, and engineering. It is also the name of a very popular conference on scientific programming with Python. The SciPy library depends on NumPy, which provides convenient and fast N-dimensional array manipulation. The SciPy library is built to work with NumPy arrays, and provides many user-friendly and efficient numerical routines such as routines for numerical integration and optimization."
i follow two useful blogs in order to learn how to plot and make basic statistical calculation by using pyton, after i learn these topics, i will search about machine learning libraries.