CSCI 300 Artificial Intelligence

gudgeonmaniacalΤεχνίτη Νοημοσύνη και Ρομποτική

23 Φεβ 2014 (πριν από 3 χρόνια και 3 μήνες)

70 εμφανίσεις

CSCI
3
00

Artificial Intelligence

Prof.
David
M.
Keil, Framingham State
University
,
Fall
20
13

SYLLABUS

Invitation

Would you like to explore the limits of computers’
ability to solve the kinds of problems that humans
solve? Would you like to discover what the
mechanisms of
cognition, learning, knowledge
, and
belief

are
; what the
mind

is
? To know how a robot
works? To

examine arguments
why
machine
s can or
cannot be conscious or intelligent
? Join us in this
course and look into the future.

Course description

An introduction to cognitive science and efforts to
implement intelligence

in computer hardware and
software
, with application areas
. Topics include
search, knowledge, reasoning, uncertainty,
adaptation, learning, embodied interaction, future
prospects, and philosophical considerations.
Concepts will be demonstrated with softwar
e.

Prerequisites:
Junior standing, CSCI 258
Introduction to Operating Systems Using

UNIX
TM
,
and either MATH 117 Introduction to Statistics or
MATH 200 Precalculus.
.

Strongly recommended
reading

S.
Russell and
P.
Norvig,
Artificial Intelligence:
A

Modern
Approach
,
3
r
d

ed., Prentice Hall, 2010
.

Handout material
,

including
s
lides
: see
course site.

I like this textbook because it’s readable and up to
date.
Learning about AI will involve r
eading and
studying text material related to the course.
Students
are in
vited to find
alternative

text material of
comparable value on their own.

Meeting
times

Tuesday 6:30
-
9:5
0

p.m.

Hemenway Hall
132

(annex)

To contact
me
:

Office hours
(Hemenway Hall 318A):

Tues., 10:30
-
11:30 a.m., 5:30
-
6:30 p.m.

Thurs. 10:30
-
11:30 a.m.

O
thers by appointment

Office:
Hemenway Hall 318A

Telephone:
(508) 626
-
4724

Email:
dkeil@fr
amingham
.edu

URL:
www.framingham.edu/~dkeil

Course overview

Other possible names for the field of artificial
intelligence are “Solving hard computational
problems” and “Rational adapti
ve computational
behavior.” By
hard

problems
, we mean
ones
subject
to combinatorial explosion in the number of possible
solutions to consider, or
interactive

problems that
require l
earning and adaptation. In this course, we
emphasize interaction and the notion of
rational
agents


computational entities that take good
actions
in response to
their
percepts (
inputs
)
.

Our interest includes
models

of intelligent
agents

and of
systems
cap
able of intelligent behavior.
Intelligence requires
adaptation

to an
environment
via
behavior

that is designed to change the
environment.

Intelligent behavior is most typically
interactive

and is often,
social

behavior of systems
that comprise
multiple act
ors
.

Intelligent behavior is
often
associated with
algorithmic

problems that involve planning several
interaction steps ahead, so that the number of
possible situations to consider explodes

the farther
ahead we look
. The task for the intelligent agent is t
o
find a approximately optimal action among a huge
number of possible actions

To
handle
many
intractable
optim
iza
tion

problems,
an approach of
bounded rationality

is suggested, so
that
satisficing

rather than
optimality

is the goal.
Intelligent behavior, therefore, is an approx
imation
of perfect behavior; it is “doing the right thing” in the
sense of getting close enough to exactly the right
thing. It is the time constraints of real
-
time
interaction that forces the reso
rt to satisficing.

AI seeks to model
cognition
, including knowledge
acquisit
ion, storage, and application. Mathematical
t
ools include
logics

(propositional, predicate, modal
and temporal), and
probability theory
.

We will refer to the related fields of neur
oscience,
cognitive science, psychology, philosophy, control
theory, and decision theory.

D. Keil

C
SCI 3
00 Artificial Intelligence

Framingham State University

Fall 201
3

The trend of AI research has evolved from
reasoning

in deterministic
toy environments

(e.g.,
board games), to
inference

using predicate logic, to
planning
,
uncertainty
, and
belief,
to

intera
c
ting
with
an environment, to

coordination
of multiple agents in
dynamic environments.

What we will investigate

S
tudents are asked to consider the following
ques
tions and to develop their own points of view.

Is the general

pr
oblem for any cognitive system to
give timely responses to percepts from an
environment?

Can intelligence be summarized as reasoning;
ontogenetic learning

(single
-
agent); phylogenetic
learning (evolutionary); and sociogenetic learning
(distributed)?

M
ay
AI
be summarized as rat
i
onal adaptive
computational behavior
?

Timely responses to percepts from an environment
is the general problem for any
cognitive
system.

Can many problems solved using intelligence may
be reduced to the search of an
exponential
-
sized state
space?

Is it AI’s task to give approximate solutions to
NP
-
hard problems?

What are ways in which k
nowledge
is
represented
?

How can

knowledge
be acquired
?

Is
probabilistic reasoning

part of intelligence?

What is the form of knowledge

in

partially
observable and non
-
deterministic environments
?

What
type of learning
occurs in static
environments?

I
s
adaptation

a

higher
form

of intelligent behavior
than generalization
?

Are sociogenetic (distributed) and phylogenetic
(evolutionary) learni
ng higher forms than
ontogenetic (single
-
agent)?

Are situated and embodied forms of intelligence
more robust than reasoning
-
based systems?

I
s intelligence
emergent and decentralized
?

Do p
ersistent, dynamic, physical environments
pose problems solvable only

with
multi
-
agent
systems and distributed AI
?

What is the role of indirect interaction in
intelligence?

C
ould c
l
arity in research in cognitive science be
served by relaxing the association of intelligence
with

humanness and by reframing AI and cognition

as rational adaptive behavio
r
?

Does t
he notion of bounded

optimality

(a

property
of the best
program

achievable to solve a problem

that entails adaptation) offer

a
sound
theoret
ical
foundation for AI research?

Background concepts

I expect that most AI students will have learned to do
the following in Computer Science I, Statistics, or
Precalculus. We will review them and use them in
this

course.

0.1a

Explain basic precalculus concepts*

0.1b

Write the truth table for

a propositional
-
logic
formula or logic circuit*

0.2

Design a looping

algorithm*

0.3a

Find a shortest path in a graph*

0.3b

Explain the relation between the logarithm
function and the heights of trees*

0.4

Explain basic notions of combinatorics*

Learning objectives


I see
k to show that
students can:

0a.

Show knowledge of facts and concepts

0b
.

Summarize the semester’s learning

0c.

Carry out documented research on AI

0d.

Participate in class activities throughout
the

semester

0e.

Solve problems as part of a team

0f.

Present

results in the classroom

0g
.

Reflect on reading of text material

1.

Compa
re human cognition with computa
tional or
agent models of perception
-
action

2.

Explain how heuristics offer ways to pursue
goals in exponentially large search spaces

3.

Describe the representa
tion and use of knowledge
in
inference
-
based problem solving

by
knowledge
-
based agents

4.

Apply probability theory to describe
and model
agents operating in uncertain environments

5.

Describe different ways to supervise agents to
learn and improve their behavior

6.

Explain adaptive learning from the environment

7.

Explain the relation between distributed artificial
intelligence and
self
-
organized systems

8.

Defend a theory of mind, relating it to ethical
issues raised by artificial cognitive systems

D. Keil

C
SCI 3
00 Artificial Intelligence

Framingham State University

Fall 201
3

9.

(Summary) Distinguish
stages in the
development of artificial
-
intelligence research
and applications

Research paper

Each student will propose in writing a t
opic for a
short research paper; submit a preliminary version
that responds to comments about
the
proposal;
present the pa
per in class; and submit a final version
at the end of the course. Subject to proper formatting,
textual presentation, and documentation, papers will
be made available on the Web.

How the course will deliver what it offers

F
or each topic, we have presen
ta
tions, group work,
discussion, assign
ments, and quizzes. Assigned work
and quiz questions help t
o assess attainment of
learning

objectives.

Our classroom environment emphasizes active
inquiry
,

participation, respect
,

and support among all
participant
s. Learning is
seen as
the interactive
construction of k
nowledge by the learner.
We

ask

each other que
stions and investigate
problems

together
.

Work includes
small groups and blackboard work
and report backs from each student. A

research
project
encourages

student inquiry in a particular
area of interest
.

Frequent assignments and quizzes
monitor progress and
enable second chances.

Grades
assess learning
based on attainment of
the
stated

objectives

of the course
. I score each item of
work, or grading criteri
on, on a scale of 0 to 1.0.

The essay, “What we do in my classroom,”
describes exercises,

collab
oration
,
and

grading
.

Semester grading weights

The following categories group course objectives
and outcomes (see previous page), which are
assessed by means

of assignments, quizzes, exams,
and records of classroom discussion and
presentations.

Application of concepts


core topic objectives

35


other topic objectives

10

Knowledge of facts

10

Summary and reflection

5

Written contribution

15

Presenting results

in person

10

Group activity

10

Attendance

0
5



100

My r
esearch interest
s


My
research includes investigati
on of the power of
multi
-
stream, multi
-
agent, and indirect
interaction.
See
my
web site for publications related to AI

and
indirect interaction
.

Accommodations

“Students with disabilities who request accom
-
modations are to provide Documentation Confir
-
mation from the Office of Academic Support within
the first two weeks of class. Academic Support is
located in the Center for Academic Support and
Ad
vising (CASA). Please call (508)

626
-
4906 if you
have questions or if you need to schedule an
appointment.” (See
http://www.framingham.edu/

CASA/Accommodations/accomm.htm
.)






D. Keil

C
SCI 3
00 Artificial Intelligence

Framingham State University

Fall 201
3

Course Plan

Dates

Topic

Russell
-
Norvig

chapters

9/10



9/17

Introduction


Appendices A
-
B

9/17



9/17

1.

C
ognition

and
c
omputation

1
-
2
; handouts
1

9/24

2
.

State
-
space search

3
-
6

10/1

Problem
-
solving quiz

on topic

1


10/1


10/8

3
.

Knowledge
representation

and inference

7
-
1
1; 12.1
-
12.2

10/8


10/15

4
.

Uncertainty
and probabilistic reasoning

12.3
-
12.6;
13
-
17

10/22

Research reports (abstracts)


10/22

Problem
-
solving quiz
zes

on topics

2
-
3


10/22


10/29

5. Supervised learning

18
-
20; 23
-
24

10/29


11/5

6
.

Reinforcement learning and adaptation


17.5;
21
-
2
2

11/12

Research reports (preliminary drafts)

Pr
oblem
-
solving quiz on topics 4
-
5

Make
-
up
quiz
zes

on topics 1
-
3


11/12


11/19

7
.
Distributed AI and m
ulti
-
stream adaptive
interaction

12.7; 17.6; 25;
handouts
2
,
3

11/19


11/26

8. P
hilosophical c
hallenge
s
and f
uture prospects

26
-
27;

handouts
4
,
5
,
6
,
7

11/26

Research reports (final drafts)

Pr
oblem
-
solving quiz on topics 6
-
7


11/26


12/10

Summary

and review


12/3

Pr
oblem
-
solving
quiz on topic 8; make
-
ups on topics 4
-
7


12/10

Final exam part A (mul
tiple
-
topic problems)

make
-
ups on topic 8


12/17

Final
exam
part B

(
multiple choice,
all
to
pics
)

Optional questions on topic objectives






1

Thagard, pp. 3
-
15; Simon, pp. 675
-
691;
Weizenbaum pp. 128
-
131.

2

D. Keil, Indirect interaction in evolving adaptive multi
-
agent systems (2006); Decent
ralization and stigmergy (2008).


3

D. Keil and D. Goldin,
Models of Self
-
Organizing Systems: Learning from the Ants and the Bees (2007)
.

4

D. Keil, Notes for debate on AI (2005)
.

5

J. Markoff, A fight to win the future: Computers vs. humans
.
NY Times
.


6

R. Kurzweil, The coming merging of mind and machine (2002).

7

J. Searle, Minds, Brains, and Machines (1980)
.

D. Keil

C
SCI 3
00 Artificial Intelligence

Framingham State University

Fall 201
3