***********************************************************
CSCI 580 - FINAL PROJECT - Fall 2007  (40% of course grade)
***********************************************************
--last updated 3/23/08 rrenner

DUEDATES: 
 At completion of coursework, after viewing all lectures.
 Must have an approved Proposal on file before beginning.
 Must have completed the novel critique and Hwk1&2 before beginning.
 May submit BEFORE or AFTER Final Exam. 

DEMO (not applicable to offsite students, unless requested
     after submission of report, in rare cases).

PROJECT DELIVERABLES:
======================

Electronic Report:  
- 7-10 pages (appendices), double-spaced, 12pt font, 1" margins
- Minimum Inclusion of (elaboration of proposal - revisit detail in proposal specs):
o Cover page with name, title, contact info, and 1 paragraph ABSTRACT
o Introduction, including Problem Statement
o Literature Review - Related Material (journal citations)
o Methodology (replicatable detail on how you setup experiments)
o Test Plan & Results (including tables, charts, and/or graphs)
o Analysis / Interpretation of results
o Conclusions (significance, limitations, recommendations, etc.)
o Reference List (along w/citations in document, and URL to software/data)
o Appendices with Usage Information, Screenshots of testing, and
    code and/or data/results
      -OR- 
    submission of electronic data/code files


Minimum Readings Required (with in-paper citations)
===================================================
* Prechelt paper on Benchmarking (wwwipd.ira.uka.de/~prechelt)
* minimum of 1 Published journal article on topic related to your project
  (algorithm, modeling tool, method, etc.)
* minimum of 1 additional credible source for your project 
  (journal, text, proceedings, tech.report, credible web-based source, etc.)
NOTE: any references to texts, manuals, or documentation should cite not
      only the source, but the chapter/section/page number



Machine Learning Practicum:
==============================
Your project should fall into one of the following
(or an approved alternative) category:

1) design and implement a Machine Learning algorithm
   (e.g.: general nnet agent or GA agent or system of agents)

 -OR- 

2) extend an existing modeling tool (e.g. - AIMA, NNES, other...)
   with additional computational agents, search algorithms, etc.

 -OR- 

3) design and run effective experiments for benchmark testing of a
   simulation/modeling tool of choice (e.g. - AIMA, MatLab, SNNS, JNNS, 
    Neuron, Brainmaker, etc.). Experiments should be 
   well-planned, exhaustive as possible to test some 'hypothesis'
   (you determine and indicate), and duplicatable.

 -OR-

4) Robotics project, with approved platform and goals



There are four primary objectives for this assignment: 
1) to gain experience & understanding of machine learning 
   algorithms, methodologies, and/or modeling tools, 
   via hands-on activities.
2) to gain experience with standard benchmarking practices, through reading, 
   experimentation, testing, and reporting (scientific methods),
3) to hone one's writing skills by skillfully reporting on experimental 
   conditions and results, and providing a detailed analytical comparison 
   and conclusions, and
4) to experience readings in AI at the professional journal level.

Note on experiments:
Experiments may take on any one of the following (or approved) forms:

1) Intra-Simulator:  run all your experiments on one simulator, comparing
   results, where each experiment has some unique parameters, OR 
   compares to some published results for the same dataset on same simulator

2) Inter-Simulator: run your experiments on  different simulators,
   or compare your runs to published results for another simulator

3) Compare different learning or architectural algorithms/models 
   from a chosen simulator, or between two simulators.  

Use reputable data of your choice.  In your paper, detail the 
experimental conditions, the data, the results, and your assessment
of the results and why things turned out the way they did.


Sample Data Repositories:
==========================
PROBEN1: (see Prechelt site)
DELVE:  www.cs.toronto.edu/~delve/
UCI, CMU, StatLog, ELENA


Sample Simulation Tools:
=========================
AIMA, SNNS, AdaBoost, Genesis, Neuron, NeuroSolutions, 
Brainmaker, Brainsheet, NNES, BANG,...
or any other you may find that is appropriate 
(neuro, neuro-fuzzy, neuro-ga, classifiers,...)



Any Questions? email me at: renner@ecst.csuchico.edu