Department of Computer Science
University of Miami
CSC545 - Introduction to Artificial Intelligence
Spring 2008
Description
Decisions, decisions, decisions.
Solving a problem requires making decisions, and making the right decisions.
The ability to make the right decisions, and hence solve a problem, is
a fundamental measure of intelligence.
Lots of wrong decisions are made all the time - just watch old Seinfeld
reruns - so intelligence doesn't always come easy.
Artificial Intelligence is the study of computer agents that make the right
decisions (with more than random chance), and hence exhibit intelligence.
There are five cornerstones to the construction of artificially intelligent
agents:
- Powerful input processing, to obtain an
adequate description of the problem to be solved.
Examples are natural language processing, image recognition, and tactile
sensor processing.
- Problem representation, to maintain an
adequate representation of the problem as the decisions made transform
it from the initial description to a final solution.
Examples are logics, semantic nets, and Bayesian networks.
- Search strategies, to investigate alternative
decisions and to evaluate the quality of each.
Examples are A* search, iterative deepening, and intersection search.
- Contextual knowledge, to provide domain specific
information that can be used to guide the search for a solution.
The same data structures used for problem representation are
useful here.
- Powerful output processing, to present the solution.
Examples are robotic arms, speech synthesizers, and digital interfaces.
If all the above sounds kinda different to other things you've learned
in Computer Science, you're right!
Many traditional Computer Science techniques (algorithms, data structures,
etc.) are invented and refined by intelligent computer scientists, but their
execution on a computer exhibits no intelligence at all.
Artificial Intelligence aims to build computer agents that exhibit
intelligence themselves.
CSC545 - Introduction to Artificial Intelligence, provides an
introduction to some of the established computing techniques used
to build intelligent agents.
Some of the topics covered are:
- History and foundations
- Classical logic and reasoning
- Prolog programming
- Features of knowledge features
- Reasoning with uncertainty and imprecision
- State space search
- Knowledge representation
- Machine learning
Learning Objectives
- Discuss the origins and history of AI.
- Understand the structure of intelligent agents, and understand that AI
software is different from traditional computer science software.
- Translate problems from a natural language description into a logical
formalism, and solve the problem using logical inference rules.
- Understand and use the notion of state space search.
- Represent knowledge using appropriate data structures.
- Use various specialist AI techniques.
- Program in a AI programming language (Prolog this time round)
Preparation
CSC545 has the pre-requisites:
Students who do not meet the pre-requisites must tell the instructor.
Instructor
Dr Geoff Sutcliffe.
Contact details are on the WWW at
http://www.cs.miami.edu/~geoff.
The WWW page gives office hours when students are welcome, and
students are encouraged to ask questions by email at all times.
Teaching Assistant
Anoop Mohan
Contact details are on the WWW.
Contact Hours
Each week there are two 75 minute lectures, Section O:
- Tuesday, Thursday 9:30-10:45am, MM300
For some weeks there will be optional labs, for practicing Prolog
programming.
Dates and times will be negotiated to suit everyone.
Students are required to read their email regularly, and to
consult the subject WWW page regularly.
Resource materials
There is no required text - all will be revealed in the classes.
The recommended text, which covers most of the material of this course, is:
A list of
reference texts,
lecture slides,
and
assignments will be available
on the WWW.
Assessment
| 5%
| Short Essay
|
| 20%
| Prolog Programming Project
|
| 25%
| Intelligent Agent Project
|
| 10%
| Midterm Test (February 28th, in class)
|
| 40%
| Final Exam (May 7th, 8:00-10:30am)
|
In order to obtain a particular grade, you may be required to attain
that grade in all items of assessment.
Assignments will be placed on the WWW.
The submission requirements for each assignment are given with each
assignment.
Late submissions will not be accepted.
Extensions of the due date will be granted if supporting documentary
evidence is supplied (e.g., a doctor's certificate).
Application for an extension must be made to the instructor before
the due date (if possible).
Assessment items must be completed individually.
While general interaction between students is encouraged, plagiarism
is a breach of the Honor code. It is ok to talk to other
students about general solution techniques for assignments,
but it is not ok to copy solutions in part or as a whole.
Plagiarism will result in a loss of marks for all guilty students
involved.