1 Expected learning outcomes After the course students will be able - - PDF document

1

Umeå University Department of Computing Science

Emergent systems

Spring-15

Jonny Pettersson

http://www.cs.umu.se/kurser/5DV017/VT15

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15 Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Course Description

❒ Element 1, theory, 4,5 högskolepoäng

❍ The course deals with systems where the system's

behavior arises as an emergent property of interactions between system components. Emergent properties can be observed in all non-linear systems that are sufficiently complex, both natural and

• artificial. Fractals, chaos, complex systems, evolution

and adaptation are examples of topics. ❒ Element 2, assignments, 3 högskolepoäng

❍ The element consists of a number of mandatory

assignments

2

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Expected learning outcomes

❒ After the course students will be able to:

❍ exemplify how the parts of the system can interact

and give rise to a more complex behavior than is apparent in the description of each part

❍ explain how fractals, chaos, complex systems,

evolution and adaptation can be used to model, simulate and understand emergent systems

❍ compare different techniques and algorithms such as

cellular automata, ant algorithms, genetic algorithms, boids, swarm algorithms, evolutionary methods, autonomous agents, producer-/consumer dynamics

❍ explain different game theoretical models ❍ apply the algorithms and models to simulate and

visualize simple natural and artificial systems that express emergent properties

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Last time

❒ Tenth time, different approaches used ❒ An interesting course with good result ❒ Average workload ❒ The scientific approach in the

assignments are appreciated

❒ Literature

❍ Good and not so difficult to read

❒ Assignments

❍ Interesting

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

This Course

❒ Text book + papers

❍ Computer programs

❒ Focus on

❍ Fractals ❍ Chaos ❍ Complex Systems ❍ Evolution and

Adaptation

❍ Doing science

❒ Contents

❍ Lectures ❍ Assignments ❍ Project ❍ Textbook, papers

3

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Assignments and Project

❒ Assignments

❍ Termites ❍ Boids ❍ 2 and 2 in the assignments ❍ NetLogo as a tool

• First (?) contact with NetLogo

❒ Project

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Teaching

❒ Pedagogical thoughts ❒ Lecture materials

❍ Source ❍ Contents

❒ What should you learn?

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

The rest of this lecture

❒ Concepts

❍ Emergence, emergent systems, …

❒ Life ❒ Topics ❒ Scientific approach ❒ NetLogo ❒ Assignments

4

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Emergence

❒ Definitions

❍ (The whole is more then a sum of parts) ❍ (The global behavior can not be predicted

from lower levels)

❍ Something that emerges in the interactions

between (simple) parts and the environment, and that is not described in the parts ❒ Emergent properties ❒ Emergent systems ❒ Example: Ants

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Emergent Behavior

❒ Bottom-up ❒ Distributed ❒ Local determination of behavior ❒ On all levels ❒ Example: The human body

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Properties of Emergent Systems

❒ Many interactions ❒ Decentralized ❒ Non-linear ❒ Dynamic ❒ Competition and cooperation ❒ Emergent properties

5

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Many interactions

❒ In space and/or time ❒ Societies made of many people, people

made of many organs, organs made of many cells

❒ A system of parts because of

interactions

❒ Number of parts may differ ❒ Massive parallelism

❍ Often many simple parts doing the same

thing

❍ Complexity comes from interaction

❒ Example: Weather

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Decentralised

❒ Self-organization

❍ The order emerges from the system itself

❒ Advantages of decentralization

❍ Easier to adapt to changes ❍ No need for a smart leader

❒ Example:

❍ WWW

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Non-linear

❒ Do not obey the superposition principle

❍ Output is not proportional to input

❒ Interactions between parts ❒ Example: Phase transitions

❍ Solid – liquid - gas

6

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Dynamic

❒ Often the interactions continues on and on

❍ Does not always come to a ”fixed” state ❍ Can be better to handle changing environments

❒ Dynamic systems can have different amounts of

complexity

❒ Example:

Society

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Competition and Cooperation

❒ Some agents may cooperate ❒ Some agents may fight for the same

resource

❒ Some agents may destroy what others

tries to do

❒ Example: Producer-consumer systems

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Discussion

❒ Questions to consider:

❍ What is the emergent property? ❍ What is the goal of the system? ❍ Does each agent know the goal? ❍ How was the system created?

❒ Emergent systems in the society ❒ Emergent systems in the nature

7

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Complex Systems

❒ (Almost) like emergent systems

❍ Many parts ❍ Interdependent parts

❒ Difficult to understand

❍ The behavior of the whole system

understood from behavior of the parts

❍ The behavior of the parts depends on the

behavior of the whole system ❒ Example: Family

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Adaptation

❒ Can lead to improved fitness and

performance, or just to be able to survive

❒ Adaptation can happen in three ways

❍ Improved handling of an event by the agent ❍ Learning – in the lifetime of the agent ❍ Evolution – across generations

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Complex Adaptive Systems

❒ A complex adaptive system is a system

consisting of many interacting parts. The behavior of the system emerges

• ut of the parallel interactions between

the parts and the environment without any global plan. The parts adapt and evolve over time.

❒ Example: Ecosystems

8

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Life – What is Life?

❒ Vitalism, - 1600

❍ Life is ”something” extra over and above the detailed

• rganization of a material organism

❒ Langton, 1988

❍ ”…living organisms are nothing more than complex

biochemical machines. … A living organism … must be viewed as a large population of relatively simple machines.”

❍ ”Life is a property of form, not matter”

❒ Flake, 1998

❍ ”Nature appears to be a hierarchy of computational

systems that are forever on the edge between computability and incomputability.”

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

From Chaos to Life

❒ Living organisms are nonlinear systems! ❒ Living organisms are complex systems! ❒ How has nature achieved this?

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

From Chaos to Life - Naturally

❒ Evolution through natural selection

❍ Darwin, The Origin of Species, November 24, 1859 ❍ Genotype – phenotype ❍ Criteria for evolution

• Heredity
• Variability
• Fecundity

❒ The role of environment ❒ Co-evolution

❍ A necessary condition?

❒ Self-similarity ❒ Self-organization

9

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

From Chaos to Life - Artificially

❒ How to do this artificially?

❍ Can not predict the global behavior of

simple interacting subparts

❍ Can not decide which subparts to use to get

a predetermined global behavior ❒ You must ”run” the system to see what

kind of global behavior it generate

❒ You need methods to search through

the solution space of a nonlinear system

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

From Chaos to Life - Artificially

❒ Methods

❍ Lindenmayer systems ❍ Cellular Automata ❍ Boids, herds and flocks ❍ Ant Algorithms ❍ Genetic Algorithms ❍ And more…

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Simulation and Modelling

❒ Emergence is explained ❒ Want the simplest system that produce

the emergent behavior

❍ Ockham’s razor (1285 – 1347/49)

❒ Not complete models of reality

❍ Focus on the essence of the system ❍ Not clones

10

Emergent Systems, Jonny Pettersson, UmU

Scientific approach

19/1 - 15

Formulate the question Communicate the model Assemble hypotheses Choose model structure Implement the model Analyze the model

(Adapted from Grimm and Railsback 2005)

Compare

Patterns Patterns

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Why Study Emergent Systems?

❒ Fundamental to theory and

implementation of massively parallel, distributed computation systems

❒ Goals/challenges

❍ Efficiency ❍ Self-optimizing ❍ Adaptive ❍ Robust to failures ❍ Security

❒ Nature as a source

❍ Try to understand nature ❍ Use nature as an inspiration

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Assignments

❒ NetLogo

❍ A multi-agent modeling language ❍ A parallel extension of Logo

❒ Rules for assignments

❍ See the links on the course homepage

❒ Assignments

❍ Termites ❍ Boids

11

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Assignments – The reports

❒ Focus on reflections and understanding ❒ The report should have a scientific

approach

❍ Background ❍ Method ❍ Result ❍ Discussion

❒ You may cooperate on code but not on

reflections and report

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Summary

❒ Concepts

❍ Emergence, emergent systems, …

❒ Life ❒ Topics ❒ Scientific approach ❒ NetLogo ❒ Assignments

Emergent Systems, Jonny Pettersson, UmU 19/1 - 15

Next time

❒ Fractals ❒ More on the assignments and NetLogo