Cognitive Systems Experts Group Seminar Adaptive Learning Systems Applications

1
Cognitive Systems Experts Group
Seminar-Adaptive Learning Systems Applications
Department of Cybernetics
PRESENTED BY Dr Will Browne
w.browne_at_cyber.reading.ac.uk
2
Contents

• Why Use Learning Systems?
• What Are Learning Systems?
• Example Applications
• The Future

3
Why Use Learning Systems?

• 1. To solve interesting problems.
• A travelling salesman must visit every city in
  their territory exactly once, and then return
  home covering the shortest distance.
• Search space S is S (n - 1)!/2.
• A 10-city problem has 181,000 routes
• A 20-city problem has 10,000,000,000 ,000,000
  routes

4
Why Use Computers?

• 1. Search space
• (enumeration may take millions of years!)
• 2. Complexity
• (simplification renders the solution useless)
• 3. Objective function
• (noise and time variance)
• 4. Constraints
• (difficult to find one feasible solution,
• harder to find optimum solution)

5
Where to Use?

• Numerical, Combinatorial Optimisation
• System Modelling and Identification
• Planning and Control
• Engineering Design
• Data Mining
• Machine Learning
• Artificial Life

6
Benefits

• Procter Gambles use of agent-based computer
  models saves 300 million annually.
• (Source Computerworld.com)
• Siemens AG realise savings of 100,000-140,000
  on each optimised heat exchanger network.
• (Source Nutechsolutions.com)
• U.S. government tax agency recouped millions in
  revenue
• (Source SPSS.com)

7
Theoretical Division of AI
KNOWLEDGE BASED
ENUMERATIVES
GUIDED
NON-GUIDED
Expert
Decision
Case Based
Backtracking
Branch
Dynamic
Systems
Support
Reasoning
Bound
Programming
INTELLIGENT AGENTS
(inc. Artificial Life)
FUZZY LOGIC
IMMUNE
CELLULAR
ANT
SYSTEMS
AUTOMATA
COLONY
LEARNING
GUIDED
NON-GUIDED
Tabu
Search
Las Vegas
Simulated
Annealing
GENETIC EVOLUTIONARY COMPUTATION
NEURAL NETWORKS
Hopfiled
Kohonen
Multilayer
Maps
Perceptrons
GENETIC ALGORITHMS
GENETIC
PROGRAMMING
EVOLUTION STRATEGIES
PROGRAMMING
8
How to Learn

• Feedback from the environment
• Supervised Learning
• Teacher provides correct response to be learnt
• Do not lend to this type of person
• 2. Reinforcement Learning
• Teacher provides reward to direct learning
• Lending to this type of person cost x
• 3. Unsupervised Learning
• Teacher is internal to the learning system.
• This person is close to this type of person

9
Evolutionary Learning

• EC
• EP
• ES
• GA
• GP
• LCS

Evolutionary Computation Evolutionary
Programming (Fogal 1962) Evolutionary
Strategies (Rechenberg 1973) Genetic
Algorithms (Holland 1975) Genetic
Programming (Koza 1992) Learning Classifier
Systems (Wilson 1994)
10
Industry

• Communication mobile phone ground station
  satellite networks
• Scheduling of work, transport, crane operations
  and so on
• Routing of computer networks.

INTELSAT operates a fleet of 19 satellites
11
Engineering

• Optimisation of route planning
• Design of complex structures
• Process optimisation

12
Control

• Domestic appliances, such as Microwave ovens
• Traffic flows
• Aircraft flight manoeuvres

13
Academia

• Game playing, e.g., chess
• Robotic football
• Test problems, e.g., iterated prisoners dilemma.

14
The Future

• 105 Data Mining Companies
• (Source KDNuggets)
• Rationalisation of companies
•    Amdocs Acquires Xchange for 5.1 Million in
  Bankruptcy Auction
• Beneficial problems exist requiring techniques to
  provide solutions
• Continue developing and creating techniques
• Match techniques to problem domains

15
Summary

• Learning based on feedback and biological
  metaphors
• Great practical potential and application
• Popular in many fields
• Yields powerful and diverse solutions
• High-performance with low costs