Homework 0 IRIS FLOWERS

CS5785 Homework 0
OVERVIEW
 A complete submission should
include:
1. A write-up as a single .pdf file
2. Source code and data files for all of your experiments (AND figures) in .py files if you use
Python or .ipynb files if you use the IPython Notebook. If you use some other language, include all build scripts necessary to build and run your project along with instructions on how
to compile and run your code.
The write-up should be in professional lab report format. It should contain a general summary of
what you did, how well your solution works, any insights you found, etc. 
1 SUBMITTING HOMEWORK
All homework must be submitted via CMS, Cornell’s course management system. Instructions for using
CMS can be found on this semester’s Piazza page, https://piazza.com/cornell/fall2016/cs5785/
home.
We encourage Piazza for all homework-related discussion. If you have a question, please do not E-mail
the TAs directly. Rather, post your question to Piazza so all students can benefit!
1. Enroll on Piazza and read the CMS submission instructions.
2. Please read and follow the formatting guidelines on the preceding page while preparing your
homework.
2 SETTING UP PYTHON
1. Find your teammate. You are encouraged (but not required) to work in groups of 2. If you do
decide to work in a team, inlcude both teammates’ names on the report and in your submission
email. One good way of finding teammates is to check the “Search for Teammates!” post on Piazza.
2. Set up a working environment for machine learning. You are free to use whatever language you
prefer, but the TAs will only support Python.
To set up Python on your machine, download the “Anaconda” distribution, which includes Python
and several libraries that we will use in class. It is available for Windows, Linux, and Mac OS X here:
http://continuum.io/downloads
After installing Anaconda, you can develop in Python in one of many ways:
• Anaconda includes the “IPython Notebook,” which provides a web-based interface to write
and run Python code. Data scientists like IPython Notebook because a notebook can include descriptions, comments, graphs, and figures in-line with the source code that generated them.
To start the IPython Notebook, run the Anaconda “Launcher” and click “IPython Notebook.”
This will open up the Notebook server in a terminal and a web browser. You can then create
a notebook and add Python code.
• If you instead prefer to develop in a traditional more familiar IDE, Anaconda includes “Spyder”. There are also instructions on how to point Eclipse to your new Python installation here:
http://docs.continuum.io/anaconda/ide_integration
3 IRIS FLOWERS
In 1935, Edgar Anderson went to his favourite pasture and recorded the length and width of the sepals
and petals on several flowers in the field. For whatever reason, this dataset became one of the oldest and
most well-known “sanity-check” datasets around, being cited by countless papers. This class continues
this time-honored tradition by using Iris Flowers to sanity-check your Python environment and plotting
libraries.
2
CS5785 Fall 2016: Homework 0 Page 3
1. Find and download the Iris Flowers dataset from the UC Irvine Machine Learning datasets archive
at http://archive.ics.uci.edu/ml/datasets.html Hint: The iris.names file describes the
structure of the dataset. How many features/attributes are there per sample? How many different
species are there, and how many samples of each species did Anderson record?
2. Figure out how to parse the dataset you downloaded. Load the samples into an N ×p array, where
N is the number of samples and p is the number of attributes per sample. Aditionally, create a
N-dimensional vector containing each sample’s label (species).
Hint: Python has a built-in CSV parser in the csv library, or you can use the "string".split(...)
method.
Hint 2: Here is some code that prints each line in a file:
for line in open ("/ path /to/ filename .txt") :
print " Line contains : "+ line
3. To visualize this dataset, we would have to build a p-dimensional scatterplot. Unfortunately, we
only have 2D displays so we must reduce the dataset’s dimensionality. The easiest way to view the
set is to plot two attributes of the data against one another and repeat for each pair of attributes.
Create every possible scatterplot from all pairs of two attributes. (For example, one scatterplot
would graph petal length vs sepal width, another would graph petal length vs. sepal length, and
so on). Within each scatterplot, the color of each dot should correspond with the sample species.
Ideally, we’re looking for something like this figure from Wikipedia:
But your results do not have to be this ornate. Presenting six separate figures in your report is
certainly fine. Be sure to include the source code for all plots!
3
CS5785 Fall 2016: Homework 0 Page 4
Hint: This is one way to draw a scatterplot. Use whatever works for you.
from matplotlib import pylab as plt
import numpy
xs = numpy . array ([1 , 2 , 3 , 4 , 5 , 6 , 7])
ys = numpy . array ([3 , 2 , 5 , 1 , 3 , 3 , 2])
colors = ["r","r","r","b","b","g","g"]
plt . scatter ( xs , ys , c = colors )
plt . savefig (" plot .png")
Hint: If you would like plots to appear right inside of your IPython Notebook, restart the kernel
and evaluate the following before running anything else:
% pylab inline