CSC 104 Tutorial Notes
Bowen Hui
August 03rd 2005


General Questions
=================
- open questions


AI and Machine Learning
=======================
1. Assume that an interface agent observes actions made by a user and
creates (learns) rules, which can be used in the future to assist the
user. The agent has currently learned the following rule:

IF (user selects menu X) AND (X contains option "Print") THEN 
(user selects option "Print")

- why/when is this rule useful?
  - creating shortcuts
  - predicting user actions, help completion, etc.
  - shortening the menu bars by using ">>"

Imagine now that at some point the user selects a menu "File" 
(user selects menu "File"), which contains the option "Print" 
("File" contains "Print"), but the user finally selects the option
"Save" (user selects option "Save"). 

- what happens? 
  - it seems that the system has arrived at a contradiction

- is the original rule still valid? 
  - yes, because there might be cases of X->Print still to occur

- how does this example highlight why computers do the wrong thing?
  - pre-programmed rules are too rigid
  - pre-programmed rules need to be modified
  - not all possible cases can be pre-programmed

- how might the system accomodate this new experience?
  - ignore new experience
  - add a new rule like this:
    IF (user selects menu X) AND (X contains option "Save") THEN 
    (user selects option "Save")
    but this is now ambiguous -- after the user selects menu "File", 
    does the user want to save? or print?
    if this is the solution, it needs some additional resolution mechanism
  - add more context into the original rule:
    IF (user selects menu X) AND (X contains option "Print") AND
    (document is unmodified) THEN 
    (user selects option "Print")

Databases
=========
Assume that a CD store maintains a relational database that stores
information about CDs, customers and CD purchases. The database
contains the following tables:

Customer(CID, Name, Age, Address, )
CD (CDNumber, Title, Artist, Price)
Purchase(CID, CDNumber, Date)


- what is a primary key?
  primary key = unique identification per record (row in table)
              - one attribute or a set of attributes
  
- what is the primary key in each of the tables?
  - customer: CID
  - cd:       CDNumber
  - purchase: concat(CID,CDNumber) both must be used

- what does the following SQL query return?

SELECT Customer.Name
FROM Customer, Purchase
WHERE Customer.CID = Purchase.CID AND Purchase.Date = "05/08/2005"

  "show me the customer's name who purchased a CD on august 05th 2005"

- why does it have to be "FROM Customer, Purchase", 
  and not just "FROM Purchase"? 
  - because the .Name field is only available in the customer's table,
    yet we have a constraint on the date of purchase


Recall that the WHERE part of the SQL query contains a condition
(Boolean expression). 
Aside: the syntax used in JavaScript is not the same as that of SQL.

          JavaScript: BoolExp              SQL: BoolExp
Meaning   condition                        constraint
          var_x == var_y                   var_x = var_y
          BoolExp && BoolExp               BoolExp AND BoolExp
          BoolExp || BoolExp               BoolExp OR BoolExp


- write the following in SQL:
  "get all the CD titles whose price is either greater than $30 
  or less than $10"

step 1. write down all 3 parts
SELECT 
FROM 
WHERE 

step 2. decide if info pertains to 1 table, or more.
all info can be found in: CD (CDNumber, Title, Artist, Price)
SELECT 
FROM CD
WHERE 

step 3. fill in which attribute to SELECT in table
SELECT Title
FROM CD
WHERE 

step 4. identify constraints on attribute
SELECT Title
FROM CD
WHERE Price ... OR Price ...

step 5. final query
SELECT Title
FROM CD
WHERE Price > 30 OR Price < 10