CSE 150 - Week 10 Discussion Notes - Final Project, Final Review

CSE 150 – Introduction to Artificial Intelligence
Week 10 –Discussion Notes: The 'Final' Section
November 17, 2004

(Note: The Friday (Dec 3) Section has been moved to Thursday (Dec 2) from 2:00 -2:50 in Center 218)

Outline:
1: Comments on Assignment 3: Machine Learning
2: Assignment 4:Otter Programming - Submission
3: Final Review

1: Comments on Assignment 3: Machine Learning

The quality of the Assignment 3 reports far exceeded that of the reports for the first two assingments. In general they were well organized, contrained interesting content, and made good use of equations, diagrams, and tables when necessary. The following comments and suggestions are for your benefit:

1) Be professional - Many reports used narritive to express ideas. Some narritive can be long winded and distracting. Try being concise using only relevant metaphor and experiences that aid the points you are making.

2) Justify your statements with statistics - The justification for the use of many features was based on inituition rather than empircal facts. Use your intuition to develop tests that can confirm (or refute) your intuition.

3) Avoid verbal garbage - Many introductions included verbal garbage like "torrent amount of SPAM out there today". This can push away a reader. Also, there is often a value judgement associated with such comments. You are a scientist that should focus on fact, not personal opinions. Let your read form his or her opinion for themselves based on what you present.

4.) Focus on the question at hand -When answering the "Strengths/Weaknesses of NB Classifier" question, many individuals focused on the importance of good feature design and training data set. These are general issue that are relavent to all machine learning algorithms. Feeding a good algorithm bad input will produce bad output. This is obvious. The strengths and weaknesses of naive Bayes classifier are those that are associated with the naive Bayes assumption.

5) Clearly state what your results mean - Training set error and Test set error are very different. Make sure you tell the reader exactly what you are reporting.

(Note: 'Grades' on Webct should now be working. I hope to have the grades posted my Thursday night. Please check Webct and email me if you are having trouble accessing your grades.)

2: Assignment 4:Otter Programming - Open Issues

The purpose of Assignment 4 is to learn about FO logic and to gain experience with a Theorem Proving Language. Your report will make up a larger portion of the grade since the code submission requirement are significantly less for this assignment than the previous assignments. Focus on the questions that are asked in the problem statement.

You should hand in one otter input file that contains the A) Knowledge Base for the Yale Shooting Problem, and B) some queries sentences that are commented out. The program should be structured as follows:

---------------- myYSP.in -----------------------
%Name
%Comments
%%% Knowledge Base for Yale Shooting Problem %%%

Insert all axiom here
    - equality and inequality axioms
    - Unique Names Assumption for Constant Symbols
    - Causes, Cancels, and Holds Predicates
    - Narrative - Initial State, Sequence of Actions

%%% Queries %%%%%%%%%%%%%%%%%%%%%
%Query 1: Is the turkey dead after it has been shoot
-Holds(isDead, s3).

%Query 2: Is the gun loaded...
%Holds(isLoaded, s1).
---------------- end of file ----------------------

I will test your code by uncommenting each of your queries and by adding one or two of my own. Your code should be well commented so that I can add these queries.

3: Final Review

The Final exam is on Monday, December 6th form 7-10pm. The format will be similar to that of the Midterm, but longer. There will be 3 long Part 1 questions and about 25 short true/false questions. All rules from the midterm apply. You have two past midterm and two midterm solution guides to help you study. Reviewing the course lecture notes will also be particularly useful.

The three major topics for the course include: Search, Probability/Machine Learning, and FO logic.

Today we will focus on First Order Logic. (See Week 6 notes for a brief review of the other topics)

Concepts to study:
    Definitions - Logic, Knowledge Base, Expert System, Ontology
    Syntax and Semantics
    Propositional (Zero-Order) vs. Predicate Logic(First Order)
    Symbols - Constants, Functions, Predicates - How are they similar/different? - How are they "uninterpreted"?
    Variables and Quantifier - :Loves" example from week 8 notes
    Situation Calculus - fluents, actions, states
         Frame Problem - Holds, Causes, Cancels predicates
    Entails - FO Logic decidability and semi-decidability - Godel's Completeness and Incompleteness Theorems.

We will also look at questions 9-15 of last years midterm to get the flavor for Logic True/False Questions.