Sunday, June 16, 2013

2D Histrograms with Plotly

Plotly is an online tool that makes us able to create wonderful interactive visualizations of our data. It can plot data from csv files, spreadsheet, etc. but it also has a Python sandbox where we can put our Python snippets! In this post we will see a simple example that shows how to plot a 2D histogram in Plotly.

First, we need a snippet to generate some random sets of data:
from numpy import *
 
# generate some random sets of data
y0 = random.randn(100)/5. + 0.5 
x0 = random.randn(100)/5. + 0.5 
 
y1 = random.rayleigh(size=20)/7. + 0.1
x1 = random.rayleigh(size=20)/8. + 1.1
 
y2 = random.randn(50)/10. + 0.9
x2 = random.rayleigh(size=50)/10. + 0.1
 
y3 = random.randn(50)/8. + 0.1
x3 = random.randn(50)/8. + 0.1
 
y = concatenate([y0,y1,y2,y3])
x = concatenate([x0,x1,x2,x3])

The distribution of the variable x looks like:


The distribution of the variable y looks like: And the 2D histogram of both variables looks like this:


As showed in the colorbar, cells with lighter colors correspond to high density areas of the our distribution.

All the plots above were made with Plotly inside their Python sandbox using the following code:
## place the data into Plotly's dict format

# histograms
histx = {'x': x, 'type':'histogramx'}
histy = {'y': y, 'type':'histogramy'}
hist2d = {'x': x, 'y': y, 'type':'histogram2d'}

# scatter plots above the 1D histograms
# "jitter" the scatter plot points to make their distribution easier to distinguish
jitterx = {'x': x, 'y': 60+3*random.rand((len(x))), 'type':'scatter','mode':'markers','marker':{'size':4,'opacity':0.5,'symbol':'square'}}

jittery = {'x': y, 'y': 35+3*random.rand((len(x))), 'type':'scatter','mode':'markers','marker':{'size':4,'opacity':0.5,'symbol':'square'}}

# scatter points in the 2D histogram
xy = {'x': x, 'y': y, 'type':'scatter','mode':'markers','marker':{'size':5,'opacity':0.5,'symbol':'square'}}

# NOTE: the following lines plot all the graph above
plot([histx, jitterx], layout={'title': 'Distribution of Variable 1'})
plot([histy, jittery], layout={'title': 'Distribution of Variable 2'})
plot([hist2d,xy], layout={'title': 'Distribution of Variable 1 and Variable 2'})
Plots made with Plotly automatically provide interactions (click-drag to zoom, double-click to autoscale, shift-click to pan) and are very easy to embed in web page using the embedding snippet.

Thanks to the Plotly guys for providing the code of this post and this amazing tool :)

Saturday, May 25, 2013

A colorful Trifoil Knot

Some time ago I was looking for a cool way to visualize parametric equations and I started working on the the trefoil knot. It is the simplest example of a nontrivial knot and it can be defined with the following parametric equations:


After a while this snippet came out:
from numpy import linspace,pi,cos,sin
phi = linspace(0,2*pi,100)
x = sin(phi)+2*sin(2*phi)
y = cos(phi)-2*cos(2*phi)
z = -sin(3*phi)

from pylab import scatter,subplot,cm,show
from numpy import abs
s=abs(((y+3)+2)*10)
scatter(x,y,c=x,s=s,edgecolor='w',cmap=cm.gist_rainbow)
scatter(x/1.4,y/1.2,c=x,s=s,edgecolor='w',cmap=cm.gist_rainbow_r)
scatter(x/1.8,y/1.5,c=x,s=s,edgecolor='w',cmap=cm.gist_rainbow)
show()
So, here's my colorful version of the trifold knot:

Friday, May 3, 2013

A new RefCard from the GlowingPython!

Check out the DZone RefCard from the GlowingPython:


This Refcard is a collection of code examples that introduces the reader to the principal Data Mining tasks using Python. In the RefCard you will find the following contents:
  • How to import and visualize data.
  • How to classify and cluster data.
  • How to discover relationships in the data using regression and correlation measures.
  • How to reduce the dimensionality of the data in order to compress and visualize the information it brings.
  • How to analyze structured data with networkx.
Each topic is covered with code examples based on four of the major Python libraries for data analysis and manipulation: numpy, matplotlib,sklearn and networkx. Here is a preview of the first two pages:


Click on the preview to get the RefCard!