|
|
CSC480 ARTIFICIAL INTELLIGENCE |
|
*****NEW****** ROBOPIX 2003
- Click
Here
FOR
ROBOT PROJECT - Check out all the robot resources
available to you. They can be found at:
http://www.cs.csi.cuny.edu/~imberman/ROBOTRESOURCES.htm. Hand in the Robo
Check Off Sheet .
The neural network and decision tree slides can be found
at:
http://www-2.cs.cmu.edu/afs/cs.cmu.edu/project/theo-3/www/ml.html
New
date for neural network programs. See
below.
Dr. Susan
Imberman Office Hours: Mon 8:00 - 9:00AM, 12:20 PM - 1:20 PM, Wed
2:30 - 3:30PM
SAMPLE
MIDTERM EXAM - CLICK HERE
Professor Imberman's email:
imberman@ mail.csi.cuny.edu (NO SPACES!!
The space is there to thwart any spam crawlers.)
How
to evaluate
a paper!!
Assignments
Assignment I - Part A We
will perform a modified Turing test.
You will speak to two chat bots and to
a classmate. You will email me a trainscript of your conversations. In order to make everyone as anonymous as
possible, one person/computer will be labeled person A and the other person B.
a) Talk to Maybelle - Email me transcript of conversation. The class email address for
this project is:
csiai@yahoo.com
Conversations should be no more than 20
lines. Tell me that this is Maybelle.
b) Talk
to Alice - Email me
transcript of conversation.
csiai@yahoo.com
Conversations should be no more than 20 lines. Tell me that this is Alice.
c) Check out the Turing Twinkie Test http://www.twinkiesproject.com/turing.html
d) Talk to classmate using IM or similar chat room
e) Turing paper - Computing Machinery and Intelligence
Part B
f) Classify each conversation as ALICE,
Maybelle, or human. Use the evaluation sheet.
Conversations
will be here when Part A is completed.
DUE: 2/26
Assignment 2
Read
and evaluate the Turing paper and the intelligent agent
paper given in class according to
the rules for evaluating
a paper .
DUE 2/24
You
may also be interested in the CROSSROADS
home page. CROSSROADS is a magazine
written by student members of the ACM (Association for Computing Machinery).
Assignment 3 DUE March 12
Using one uninformed search method as defined in class, solve
the cannibals and missionaries problem.
Extra Credit - Due
March 26 Successful completion earns you 5/5 points for Assignment 3 plus 2
percentage points added to your grade.
Write a computer program, in the language of your choice that uses one
of the uninformed search methods defined in class to solve the cannibals and
missionaries problem.
Assignment 4 - Due March 26
Solve 16 puzzle (given in class) Using A* search or greedy search with heuristic h2 as
described for the 8 puzzle in Chapter 4 of your text. The heuristic h2 is the
Manhattan distance each tile needs in order to move to its correct position. Show each configuration generated by the
algorithm and indicate which configuration is chosen for expansion.
Initial Configuration:
5 1 2 4
9 6
3 8
blank 10 7 11
13 14 15 12
Goal
1
2 3 4
5
6 7 8
9
10 11 12
13
14 15 blank
Extra Credit - DUE 4/15 Write a program that solves the 16
puzzle using either greedy search or A* search.
Extra Credit - DUE 4/15
Genetic Algorithms. Solve the 8
queens problem using a genetic algorithm.
2 percentage points added to your grade.
Minimax Homework - Algorithm for Minimax with
Alpha-Beta Pruning
Do minimax and minimax with alpha beta pruning on the
following two problems. Problem1 Problem2
Extra Credit DUE 4/28
Worth (2 points) - Write a program, in the language of your
choosing, that plays tic tac toe (man/woman against computer) using minimax with
alpha beta cutoff or straight minimax.
Use the Evaluation function discussed in class (my possible winning
moves - my opponents possible winning moves)
Assignment
5 DUE 4/15
Evaluate
the Deep Blue paper according to the rules for evaluating a
paper .
Do you think that Deep Blue had an unfair advantage? Use facts from the paper to support your
opinion. You may also use information
from deep blue's web site at:
http://www.research.ibm.com/deepblue/home/html/b.html
Assignment 6 DUE 4/9 Show the calculations and
the resulting decision tree given by the ID3 algorithm when given the following
training data. The target concept is giveloan
|
income |
credit
card |
ownshome |
married |
giveloan |
|
<30 |
yes |
yes |
yes |
yes |
|
30-80 |
yes |
no |
no |
no |
|
>80 |
yes |
no |
no |
yes |
|
<30 |
no |
no |
yes |
no |
|
>80 |
no |
no |
yes |
yes |
|
>80 |
no |
no |
yes |
yes |
|
30-80 |
no |
yes |
yes |
yes |
|
30-80 |
yes |
no |
no |
no |
|
30-80 |
yes |
no |
no |
no |
|
<30 |
yes |
no |
yes |
no |
|
30-80 |
no |
yes |
yes |
yes |
What
rules are defined by the decision tree??
Homework - Not collected, but do it anyway!!
Neural
Nets and Decision Tree Handout Do decision tree problem. This was an old test question!!
Extra credit (1 percentage point added to grade) - Do either A or B. Do not do both. You will receive credit only for one of them.
Choice
A - C4.5 is a decision tree algorithm that implements the methods we
discussed. Download and run C4.5 on the
play tennis data discussed in class and the restaurant data on page 534 in your
text. Print and hand in the
results. C4.5 can be found at: http://www.cse.unsw.edu.au/~quinlan/ Note:
C4.5 is a Unix program and you will need a Unix account to run the software. If you are interested in this project see me
so I can give you an account if you
don't already have one.
Choice
B - The commercial windows version of
C4.5, called See5, is available for download at: http://www.rulequest.com/download.html It is limited version of the full
software. This means that it will not
work with data that has more than 200 records.
Notwithstanding, this is sufficient for the play tennis and the
restaurant data. Download See5 and
print the results from those two datasets
Assignment 7 Due April 30: Write a
program that implements the back propagation algorithm discussed in class. Have your code implement a neural network
with 2 input nodes, 2 hidden nodes, and 2 output nodes. You will need this code to program your
robots. Each team member must have a
running program. To help you check out http://www.generation5.org/cbpnet.shtml
. There is an explanation, with source
code for back propagation. The neural
net here will train a neural net to implement the XOR function. The source code is downloadable and will
run. You can modify this code for your
assignment.