Useful Techniques in Artificial Intelligence: Resented By: ILL Rowne
Useful Techniques in Artificial Intelligence: Resented By: ILL Rowne
Useful Techniques in Artificial Intelligence: Resented By: ILL Rowne
Useful Techniques in
Artificial Intelligence
-
Introduction
Cybernetics,
University of Reading
Whiteknights
Reading
UK
Picture of Lt Commander Data
This 1100 spin Bosch machine is incredibly
quiet and positively high-end. It has
everything you would expect to find on a
Bosch including exclusive features like
the 3D AquaSpa wash system with Fuzzy
Control.
Stanley
http://en.wikipedia.org/wiki/Darpa_grand_challenge
http://www.darpa.mil/grandchallenge/index.asp
http://www.darpa.mil/grandchallenge/gallery.asp
Aim
To introduce the field of artificial
intelligence,
so that it is possible to
and be able to
• Applications of Techniques
• Summary
Finance & Business
• Fraud detection
Industry
• Process optimisation
Control
• Traffic flows
• Robotic football
Artificial :-
easily understood
Artificial Intelligence :-
whole concept can be discussed
Intelligence :-
easy to recognise
hard to define
Artificial
• Computer-based
(workstation, PC, parallel-computer
or Mac)
• Computer programs
Artificial Intelligence
• Assembler
(stochastic)
Search Optimisation
Modelling
Knowledge-handling
Routing Scheduling
Visualisation Design Querying
Learning
Game-playing Adaptive-Control
Rule-Induction
Data-Access Data-Manipulation
PredictionDiagnosis
Function Summary
EXPLORE v EXPLOIT
Modelling -- Explore
Knowledge-Based -- Exploit
NON-GUIDED GUIDED
Expert Decision Case Based Backtracking Dynamic Branch &
Systems Support Reasoning Programming Bound
INTELLIGENT AGENTS
FUZZY LOGIC (inc. Artificial Life)
LEARNING
ANT CELLULAR IMMUNE
GUIDED COLONY AUTOMATA SYSTEMS
HILL CLIMBING
Tabu
REINFORCEMENT LEARNING
Search Simulated NON-GUIDED
Annealing
Age > 25
T F
Sex = F
T F T F
250 300 300 425
Fuzzy Logic
Distribution
in
department F M
Detergent :
Water ratio
Silk Wool
2 4 6 8
Weight
Learning:
Guided Search
What: Optimisation techniques that avoid
being trapped in local optima.
How:
1. Population of solutions created
2. Fitness of each solution evaluated
3. Best solutions mated for new
population
4. Repeated until optimum solution.
Genetic Algorithms:
Optimise numeric solution of fitness
function.
Learning Classifier Systems:
Optimise the co-operation of rules for
solving and input/output thickness
function.
Genetic Programming:
Optimise the interaction of code to
solve a programming function.
Evolutionary Systems:
Optimise the solution based on a
behavioural (phenotypic) instead of
genetic (genotypic) level.
F(x) = cos(x) + sin(x2) : 1 < x< 3
2
1.5
0.5
0
1 1.5 2 2.5 3
-0.5
-1
-1.5
-2
GA: j1 = 00010001
j2 = 01110001
j3 = 10010101
CONNECTION
INPUTS OUTPUTS
Representation - Required
transparency
[Wilson 97]
Interpolate & Extrapolate
• Aliasing
1.2
0.8
x x Learnt
0.6 Actual
0.4
0.2
0 x x
0 1 2 3
-0.2
• Incomplete picture
0
-0.2 0.7 1.2 1.7 2.2 2.7
-0.4
xx
-0.6 xx
-0.8
xx
-1 x
-1.2
-1.4
-1.6
-1.8
-2
Garbage In = Garbage Out
Verify
Validate
Test
Lack of Transparency