Transcript Machine Learning 1 COMP 307 30 Aug 2005

Introduction to Artificial Intelligence
COMP 307
Xiaoying Sharon Gao
Mengjie Zhang
Computer Science
Victoria University of Wellington
The 307 Group
COMP 307 1:2
• Lecturers
• Xiaoying Sharon Gao, CO229, 463 5978, [email protected]
• Mengjie Zhang, CO355, 463 5654, [email protected]
• Tutors
• Urvesh Bhowan, [email protected]
• Bing Xue, [email protected]
• Su Nguyen, [email protected]
Menu
• Course Organisation.
• What is Artificial Intelligence?
• Tasks for Artificial Intelligence
• Approaches to Artificial Intelligence
• History of Artificial Intelligence
• Reading: text book Ch 1, 2, 26, 27
COMP 307 1:3
The Course
COMP 307 1:4
Objectives:
• Understand the basic problems, principles, approaches, and
algorithms used in AI
• Be able to use a variety of AI techniques to solve real
problems
• Have a basis for further learning and research in AI
Lectures, tutorials, helpdesks
COMP 307 1:5
• Lectures:
• Week 1: Monday Friday 3:10-4:00, HT119
• Other weeks: Monday, Tuesday 3:10-4:00, HT119
• Tutorials
• Some weeks after week 1
• Friday 3:10-4:00, HT119
• Helpdesks:
• Some weeks after week 1
• Lab/date/time to be announced
• tutorials and helpdesks will be announced in lectures, on web
page or by email.
• First week: no tutorials, no helpdesks, no Tuesday lecture.
Course Outline: Topics
COMP 307 1:6
• Prolog
(1.5 weeks)
• Rule based systems
(1.5 weeks)
• Machine learning
(3 weeks)
• Evolutionary computation
(2 weeks)
• Natural language understanding
(1.5 weeks)
• Other
(2 weeks)
• Search Techniques
• Clustering or classification
Assessment
COMP 307 1:7
4 assignments
•
•
•
•
•
Hand out in lectures
Mostly due 2-3 weeks later
each about 5-10%, total of 25%
mixture of programming and discussion/analysis/paper exercises
submit online, sometimes on paper (to be announced)
Exam
• 3hr
• 75%
Mandatory requirement:
• at least a D grade on exam
Course Materials
COMP 307 1:8
• Text book:
• Stuart J. Russell and Peter Norvig, Artificial Intelligence. A Modern
Approach, Prentice-Hall, NJ, 2nd edition, 2002 or 3rd edition, 2009
• Prolog: online manual, online tutorials, some books in library.
• Web page
http://ecs.victoria.ac.nz/Courses/COMP307_2012T1/
Read course outline (if it is ready)
What is AI?
COMP 307 1:9
•
Programming computers to solve tasks that would require
intelligence for people to solve.
•
An approach to understanding the intelligence (human or in
general) by building systems that exhibit intelligence.
•
The study of how to make computers do things which, at the
moment, people do better.
•
Computing problems we don't know how to solve yet
— the hard part of Computer Science
•
A computer passing the Turing Test.
Example Tasks for AI
• Speech recognition for ESL tutoring
• Natural language queries for search engine
• Personalised news/email filters
• Personalised Web search
• Opinion mining
• Medical diagnosis
• Specialised medical test interpretation
• Credit Card fraud detection
• Computer system configuration
• Vehicle assembly
COMP 307 1:10
Example Tasks for AI
COMP 307 1:11
• Plan daily task schedule for the Mars Rover
• Planning/Scheduling orders and deliveries from a warehouse
• Machine vision for security monitoring
• Robot landmine sweeping
• Self-driving vehicles
• Self-Customising help (the MS Office paperclip)
• Characterising gene function from experimental data
• Identifying customer categories from customer data.
COMP 307 1:12
Views of AI
Several views
Systems that think
like humans
Systems that think
Systems that act
like humans
Systems that act
intelligently
intelligently
AI: Engineering or Science?
COMP 307 1:13
Engineering:
• Building intelligent systems to solve problems in the world
⇒ Understanding mechanisms, algorithms, representations for
building intelligent systems
Science:
• Understanding nature of intelligence (human or otherwise)
⇒ Implementing models of intelligence to evaluate and understand
⇒ Exploring consequences of different algorithms and representations
Approaches to AI
COMP 307 1:14
Symbolic AI
• Representation and Reasoning at an abstract level
⇒ Representations and algorithms that manipulate symbols
The physical symbol system hypothesis: A machine manipulating
physical symbols has the necessary and sufficient means for general
intelligence. (Newell and Simon)
⇒ "Old" AI
Computational AI
• The brain doesn't have symbols; use numbers
⇒ Representation and reasoning using lower level mechanisms
⇒ Probability based models and computation
⇒ Neural Networks
⇒ Genetic and Evolutionary Algorithms
Strong AI and Weak AI
Can machine think?
Can machines be conscious?
COMP 307 1:15
History of AI
COMP 307 1:16