Computational Science Basics for Astronomers: Computational Astronomy

Abstract:

A course aimed at empowering astronomers such that they can
succesfully function in the computer rich environment of the future
and understand the language and paradigms of contemporary computer
science.

Instead of teaching a collection of tricks and computer languages, the
course would emphasize the basic computational paradigms and
algorithms that are universally useful. 
I tried to collect the absolute
minimu, or survival knowledge that

Lecture 1
   A Practical Guide To Coding Principles 
   Programming Paradigms 
   Structured Programming 
   Demonstration of The Top-Down Approach 

Lecture 2
 Testing and Debugging 
 Concepts of Object Oriented Design 
 Selected Data Structures 
    linked lists/stacks/queues

Lecture 3
  More on Data Structures
    trees/hash tables. etc
  Algorithmic Analysis 
  
Lecture 4
  KD-trees
  Searching and Sorting

Lecture 5
  More on Sorting 
   Selection/Heap/Merge/Quick Sort 

Lecture 6
  Sorting cont'd (bucket and radix sort)
  Order statistics (median, etc)

Lecture 7
  Graphs 
    Representation 
    Connectivity 
    Spanning Trees 
    Brute Force and Backtracking 
  
Lecture 8
  Mathematical insights
  Greedy Algorithms
  Algorithmic Paradigms and Complexity Theory

Lecture 9
  Artificial Intelligence
  Selected Search Algorithms from AI

Lecture 10
  Monte Carlo and related methods
  Markov Chains and MCMC

The principal language of the course is python.

If you're unfamiliar with python, the first step could be:
http://docs.python.org/tutorial/

There is an abundance of beginner tutorials (google) in addition,
choose one that you like.

For in depth knowledge (possibly after taking the course) you
can learn from this tutorial
http://greenteapress.com/thinkpython/thinkpython.html

A useful python data analysis tutorial including topics
like pyfits and matplotlib that will be useful specifically
in astronomy
http://stsdas.stsci.edu/perry/pydatatut.pdf

Programming can only be learnt by actually doing it, therefore
it would desirable for the students to apply and internalize 
the culture of computer science absorbed from the course.

To promote this, after each lecture there is an optinal homework problem. 
There are two ways to obtain grades

A) Solve 5 (i.e. 50%) of the homework problems (only solutions submitted
on time count)

====OR====

B) Instead you can demonstrate their understanding of the above concepts by
completing a project by the end of the semester. The project will
have approximately the following components:

1) retrieving some data from the internet or a database, e.g., PS1
data -optionally simulate data can be used. Writing a simple
simulation is also an option instead of retrieving data.  2) writing,
testing a program that analyses the data and uses some of the above
concepts.  Incorporation of standard libraries is encouraged (i.e.
e.g. instead of writing fft from scratch, one would use fftw, etc.)
3) running the above program possibly in scripting environment many
times with changing parameters, input, etc.  4) Display the results
graphically using a graphics package, 5) Present 6-10 slides about
the results using presentation package (e.g. powerpoint)

The idea is that the project would be a smaller, and less demanding
than a science project leading to a paper, but it should a smaller scale
simulation of the computational aspects of the full process.

Instead of scientific results, the emphasis would be on finding and
internalizing your favorite tools, and algorithms. You are expected to
identify the necessary algorithms, read and understand manuals, and
learn to use at least one tool from each category (a scripting
language, a programming language, a graphics package, and a document
authoring package, like emacs with latex, or word). In your slides
should identify where you used concepts learned in the class
(e.g. scaling, memory and CPU, linear search, depth first search,
etc).

The project will preferably be from your actual research, s.t. you can
use the software you create in the future.. E.g. I can
imagine you write a library manipulating spectra of your observations,
etc. You will have to get the project approved in advance, and if you 
cannot come up with a project, I will suggest some.

E.g. of sample projects:

--retrieve stars from a PS1 data and plot color magnitude diagrams

--calculate the two-point correlation function of SDSS galaxies as a
function of color.

--calculate skewness from simulated data

--find the C_l's from a series of MAP simulations

--fit cosmological parameters from data using MCMC

--write a Zeldovich simulation and follow the collaps of a sine-wave

--put together a simple pipeline to find stars in images, and plot
  their colors

--retrieve spectra from the SDSS database and find principal
  components