- Artificial Intelligence : Theory and Practice
- by Thomas Dean, James Allen, and Yiannis Aloimonos
- Addison Wesley, 1995
-
- 위의 책의 번역판으로
-
- 인공지능:이론 및 실제
- 김진형, 박승수, 이일병, 서정연, 백운옥 공동번역
- SciTech Media 1998년 발간
-
- 기타 유사한 내용의 교과서로서
- Artificial Intelligence: structures and strategies for complex problem solving" 2nd ed.
George F.Luger, William A. Stubblefield Benjamin/Cummings Pub. Co.
- "Artificial Intelligence" P.H. Winston Addison Wesley
- "Principles of Artificial Intelligence" Nills J. Nilson Tioga Pub. Co.
- "Introduction to Artificial Intelligence" E. Charniak & D. McDermott
- "Artificial Intelligence, a modern approach" Stuart Russel and Peter Norvig,
Prentice-Hall International Inc.
강의 계획
|
Dean's Book Chapter |
Week |
Instuctor's View on Artificial Intelligence |
Chapter 1 |
1 |
Symbolic Programming and Common Lisp :
One Hour abstract |
Chapter 2 |
2 |
Advanced Topics in
Search
|
Heuristic Search Review |
Chapter 4.1- 4.3 |
3 |
Simulated Anealing /
Genetic Algoritithm |
Chapter 4.4 |
4, 5 |
Machine Learning
|
Inductive Inference |
Chapter 5.1,5.2 |
6 |
Decision Tree |
Chapter 5.4 |
7 |
Neural Network |
Chapter 5.5 - 5.9 |
8 - 9 |
Uncertainty |
Chapter 8 |
10 - 12 |
Machine Vision Overview |
Chapter 9 |
13 |
Natural Language Understanding Overview |
Chapter 10 |
14 |
We may not be able to cover Chapter 9, 10.
Lecture Notes
- Artificial Intelligence : an Overview
- LISP : One hour summary
- Heuristic Search Review
- Advanced Search Topics (Simulated Annealing, Genetic Algorithm)
- Learning I : Inductive Reasoning, Decision Tree
- Learning II : Neural Network, Perceptron, MLP
- Learning III : Radial basis Function Network, Clustering
- Uncertainty I
- Uncertainty II
평가 방법
- 중간고사 20%
- 기말고사 20%
- 숙제 20%
- Project 30%
- Class Interaction 10%
프로젝트
내용 |
평가 모드 |
평가 지침
|
Argument on Theoritical Issues |
Paper |
Significance |
Interesting Program |
Demonstration |
Interestingness |
Survey |
Survey paper |
completeness |
조교
관련Links (copied from http://www-inst.EECS.Berkeley.EDU/~cs188/info/ailinks.html)
숙제
숙제 제출 방법
숙제 채점 결과
- Write an essay on the society of Man-Machine symbiosis. (either in English or Korean). Put it
on the class BBS.
Due September 27 11:59 PM
- Write a lisp function my-rotate which recursively shifts the element of given list. For
example, (MY-ROTATE '(A (B C) D E) 1) will return ((C B) D E A), and (MY-ROTATE '((B C D) E
F) -5) will return (E F (C D B)).
- (from Dean's book 2.7) Write a recursive function that constructs a labeled binary tree of
depth n. Label the root and its subtrees with the integer 0 through 2**n so that no two
subtrees have the same label and that, given two subtrees of different depth, the label of
the one with the greater depth is larger than the other. (Assume that the binary tree used
for the function argument is given as a nested LISP list structure.)
For 2 and 3, Turn in the source code with non-trivial example run. Due October 7 11:59 PM
- Nemo is a game played on the n by n matrix. For each row and column, attached are
ordered list of black run length. Objective of the game is to discover all the underlying
black pixels with smallest number of picks. One mis-pick is eqaul to five picks.
Develop a program to search the solution for a given configuration using the genetic
algorithm. Defind your own fitness function and explain it. Show your initial populations and
evolutions graphically. Test data set1 and set2 for your program development.
Demonstration run is required with this data1 and data2. Due October 28 11:59 PM.
- Write a lisp program for building decision tree. You may consult with the sources given in
Dean's book, but you should make node explicit so that node may store attriburtes used,
examples and children as a property. Such a construction may be simpler and easier to
debug. Make a decision tree with the data files (structure description and training
examples).
Develop a function to print out decision tree readable. (Don't spend too much time to
make tree graphically. Good indentation is enough).
A test sample will be given later. Due November 4 11:59 PM.
- Obtain or write the error-back-propagation program and train a multilayer perceptron to
contruct a handwritten digit recognizer. Data files are given as follows.
data file format, training-data
Show your neural network structure and the training parameters. Evaluate your recognizer
with the test-data. Do not turn in the source code. Due November 15, 11:59 PM.
- Write a program (in any language) for the Kohonen's self organizing feature map
algorithm. Run the program with the given data file. Use 10x10 array of neurons with your
own choice of parameters. Display the trained feature map weights in a square and
connect neighbors by straight line. Show the feature map evolution tagged by several
training cycles.
Turn in the source code and the run results. Due November 30, 11:59 PM.
It is the last homework.