Computer Science 121

1
Computer Science 121

• Scientific Computing
• Winter 2012

2
Who Am I?
Contact Google simon levy Interests Robotics /
drones, cognitive science, linguistics,
Artificial Intelligence, philosophy of mind,
programming languages, iPhone apps, ...
3
Who Are You?

• Neuroscience
• Geology
• Biology
• Undecided
• Math
• English
• Accounting

4
Who Are You?

• Senior (2012)
• Junior (2013)
• Sophomore (2014)
• Firstyear (2015)

5
How Will You Do in this Course?

• A
• A-
• B
• B
• B-
• C
• C

6
Textbook

• Based on Matlab
• Developed over 4 yaears _at_ Macalester College
• Barely in print order from http//www.lulu.com
• We'll cover Ch. 1-8 in 1st 6 weeks, then various
  remaining chapters based on interest/progress.

7
Purpose
from http//www.macalester.edu/kaplan/
cs121/policies.pdf

1. To teach you to write effective computer
   programs.
2. To introduce you to some of the main
   methodological areas of scientific computation
   and provide you with the tools you need to carry
   out computational tasks you are likely to
   encounter in research.

8
Purpose
(continued from Kaplan PDF)

• Scientific computing (???) Computing
  (programming) for the sciences
• But skills are transferable to general
  programming work
• Difficulties
• Absolute precision in specifying instructions
• Considering full range of circumstances of how a
  program will be used
• Considering how your assumptions may not be
  shared by the user
• Systematically tracking down bugs
• Breaking down complicated problems into
  simple(r) sub-problems

9
Matlab

• An integrated programming and graphical
  environment
• Interpreted interactive get answer immediately
• Also supports saving and loading large programs
• Student version available for 100 from
  http//mathworks.com
• Free, open-source equivalent Octave
• http//www.gnu.org/software/octave

10
Matlab

• Issues
• Matlab vs.
• higher-level (Excel / SPSS) package
• lower-level (C / Java / Python) language
• Matlab vs. a similar package (Mathematica,
  Maple)
• Matlab vs. Python

11
1) Abstraction

• Recipe analogy How to make chocolate-chip
  cookies?
• High-level (Excel / SPSS) version

1. Make chocolate-chip dough
2. Bake cookies

• Too abstract

12
1) Abstraction

• Low-level (Java / C) version

1. Drive to store
2. Buy dozen eggs, 1 lb. flour, 1lb. sugar, 1qt.
   milk, 1 bag chocolate chips, 1 lb. butter, 1 lb.
   lard
3. Drive home
4. Turn oven dial to 375 F.
5. Take out bowl, measuring cup
6. Measure ½ cup lard, ½ cup butter
7. Put butter, lard into bowl
8. Stir butter, lard, clockwise 20 times,
   counter-clockwise 20 times
9. Measure 1 cup flour
10. Add flour to bowl
11. Stir clockwise 10 times, ...

• Too concrete

13
1) Abstraction

• Recipe analogy How to make a cake?
• Matlab version

(http//www.skyport.com/rogue-press/chocchip/basic
.htm)

1. Mix ½ cup lard, ½ cup butter, getting them good
   and creamy.
2. Mix in 1 cup flour, ¾ cup brown sugar, ¾ cup
   white sugar, 2 eggs, 1 tsp. vanilla, and ½ tsp.
   baking soda.
3. Work this over until it is thoroughly mixed,
   then blend in 1 ¼ cup more flour.
4. Bake at 375 degrees 8-10 minutes.

14
2) Matlab vs. Mathematica / Maple

• Features
• Matlab Numerical (linear algebra)
• Maple, Mathematica Symbolic (calculus)

from http//www.physicsforums.com/archive/topic/
t45208_Matlab_vs_Maple_vs_Mathematica.html
... Matlab is excellent for manipulating data, or
dealing with any sort of matrix
calculations.. From my experience Mathematica is
ideal when it comes to symbolics such as
differentiation and integration. I've seen a few
cases where Mathematica blows Maple out of the
water when one compares the types of integrals
they can evaluate. Maple and Matlab have the
best graphics in my opinion. In both of them you
are allowed to rotate 3D graphics and zoom in on
the spot (2D) in realtime. In Mathematica things
are little more complicated, which often
elicits frustration. With Mathematica, in order
to zoom, you must change the window that you are
plotting with.
15
3) Matlab vs. Python

• Python General-purpose language popular in CS
• Interpreted
• Free, open-source
• SciPy (Matplotlib) Matlab-like packages
• Good support for parallel / high-performance
  computing