CS580 Fall 1997: Empirical Research Methods and Search, Part 1

News:

Check the Writing Center web site for tutorials and descriptions on improving your writing.
Change in timing: course will be offered one day each week: Wed 10:00-11:15
Look in ~cs580 for code for the assignment:

knapgen
A knapsack generator. if you execute knapgen without parameters it tells you how to use it.
knapgen.c
Knapgen source code
kx
Knapsacks of size x (x = 50,100,200,500,1000,2000,10000)
koptima
optimal profits for the above kx-s

We want a 3 to 5 page project report on Nov 19, 1997, describing your approach and the performance of your knapsack solver on the knapsacks k50 .. k10000.

Description:

This course will present tools and evaluation methods for experimental computer science with particular emphasis on search methodologies. The purpose is to familiarize graduate students with common programming and mathematical tools for the study of search, intelligent agents and automated search software.

CS580 is the first part of a two course sequence of two credits each. The second part, CS581, will be held in Spring 1998. The first semester consists of two focused modules: Classic Search Methods and Representation. The two modules for the second semester are: Parameter Estimation in Computer Vision and Search and Empirical Methods of Evaluation. Each instructor will be primarily responsible for one of the four modules (Bohm, Whitley, Beveridge and Howe, respectively).

Logistics

When:	Wed 10:00-11:15AM
Where:	CS Conference room, USC 111
Instructors:	Ross Beveridge, Wim Bohm, Adele Howe and Darrell Whitley
	970-491-5877, 7595, 7589, 5373
	ross@cs.colostate.edu, bohm@cs.colostate.edu, howe@cs.colostate.edu, whitley@cs.colostate.edu
	USC 237, 230, 235, 227
Office Hours:	see home pages for instructors

Course Requirements

Texts:

handouts as appropriate

Pre-requisites:

CS420 or CS540 or consent of one of the instructors. Students are expected to have a background in formal methods in computer science or AI.

Grading:

The course requires demonstration of student's grasp of the concepts as follows:

Assignments (4) 30%

Projects (2) 40%

Critique 20%

Class Participation 10%

Assignments (4)	30%
Projects (2)	40%
Critique	20%
Class Participation	10%

Logistics Related to Grading:

Each assignment must be submitted at the beginning of class on the given deadline for that assignment; late period for assignments will be the start of the next following class and will incur a penalty of 8%. Please read the departmental policy statement regarding incompletes, cheating, and class attendance at student info. We will follow the guidelines outlined in these documents.

Course Topics:

Part I (Fall Semester)

Topic 1: Classical Search Methods (Wim Bohm, instructor)
- subtopic 1a: Linear Programming
- subtopic 1b: Integer Linear Programming
- subtopic 1c: Branch and Bound
- subtopic 1d: Dynamic Programming
Topic 2: Representation and Search (Darrell Whitley, instructor)
- subtopic 2a: Characterizing nonlinear functions
- subtopic 2b: Gray Codes and Walsh Transforms
- subtopic 2c: Permutation Representations
- subtopic 2d: Characterizing search space topology

Part II (Spring Semester)

Topic 3: Parameter Estimation in Computer Vision and Search (Ross Beveridge, instructor)
- subtopic 3a: Example 1: Maximum Likelihood and Maximum a-posterior Estimation to recover object pose estimates in computer vision
- subtopic 3b: Example 2: Maximum Likelihood Estimation as a tool for characterizing non-deterministic search algorithm performance.
Topic 4: Empirical Methods of Evaluation (Adele Howe, instructor)
- subtopic 4a: basic experiment design
- subtopic 4b: statistical hypothesis testing
- subtopic 4c: statistical modeling of program behavior/performance