Academic Course: 05 Overview of Methods for Engineering Autonomic Self-Aware Systems

Overview of Methods for
Engineering Autonomic Self-Aware
Systems
Artificial Immune Systems
Designed by Mark Read

The immune system
• There are many features of the immune
system that are attractive to engineers [2]:
– Distribution and self-organisation
– Learning, adaption and memory
– Pattern recognition
– Classification
– Homeostasis

What is the immune system?
• Classic View: complex system of cellular and
molecular components having the primary
function of distinguishing self from not self and
defense against foreign organisms or substances
• Cognitive View: The immune system is a
cognitive system whose primary role is to provide
body maintenance
• Danger View : The immune system recognises
dangerous agents and not non-self

Artificial Immune Systems (AIS)
• “AIS are adaptive systems inspired by theoretical
immunology and observed immune functions,
principles and models, which are applied to
problem solving” [1]
• Several flavours of AIS have emerged based on
immunological theory:
– Clonal selection theory
– Immune network theory
– Negative selection
– Danger theory
• … but new flavours are continually emerging

Clonal Selection AIS
• Clonal selection: competition for resource
drives immune system cells to mutate and
better recognize pathogens.
• This process drives the immune response from
weakly recognising a pathogen and mounting
an ineffective response, to being highly
specific and effective.

Clonal Selection AIS
• Basic algorithm [4]:
• Clonal selection AIS are typically used for pattern
recognition and classification problems [6]

Immune Network Theory AIS
• Cells of the IS recognize not only protein
structures expressed by pathogens, but each
other also
• Through this self-recognition the IS forms
positive and negative feedback networks that
modulate the immune response [3]

Immune Network Theory AIS
• Basic algorithm [4]:
• Used in data clustering and classification applications
[6]

Negative Selection AIS
• Inspired by notions of central tolerance:
• The random generation of receptors on IS cells
allows the IS to effectively cover the space of
protein structures that fast-evolving pathogens
use (often to evade immune detection).
• Central tolerance entails deleting IS cells that
recognise proteins of the host, and thus prevents
autoimmunity.
• Though negative selection is known to be
imperfect – all healthy beings contain self-
reactive cells – it has been effectively used in
intrusion detection systems

Negative Selection AIS
Basic Algorithm [4]:
Used in intrusion detection systems

Danger Theory AIS
• Rather than detect the presence of non-host
entities, the immune system recognises
damage to tissues (danger) and associates the
damage with what is found in the tissues [5]
• If there is no damage in the tissues, immune
cells that recognise structures there are
instead suppressed

Danger Theory AIS
• Basic algorithm:
– Create pool of dendritic cells (DCs)
– Use some DCs to sample available antigen, and perceive
danger/safe signals
• Has found extensive use in anomaly detection [6]

Engineering AIS
• How to engineer an AIS for a particular
application domain [1]? :

Engineering AIS
• 1. decide on appropriate representation of the
problem’s data. This facilitates mapping the
problem onto immune concepts, which affects
how the system finds solutions.
• 2. decide on appropriate affinity measures. These
quantify how elements of the system interact
with the problem environment, and with each
other.
• 3. Select algorithm(s) to operate over immune
elements of the system

Future directions in AIS
• Impasse, better capture natural system
• Modelling/conceptual framework
• Features of problem domains

References
• [1] De Castro, L., Timmis, J. Artificial Immune Systems: A New
Computational Intelligence Approach. Springer, 2002
• [2] Read, M., Andrews, P., Timmis, J. An Introduction to Artificial
Immune Systems. The handbook of natural Computing, edited by
Grzegorz Rozenberg, Thomas Back and Joost Kok. Springer, 2011
• [3] Niels K. Jerne. Towards a network theory of the immune system.
Ann. Immunol. (Inst. Pasteur), 125C:373–389, 1974
• [4] Freitas, A., Timmis, J. - Revisiting the Foundations of Artificial
Immune Systems for Data Mining. IEEE Transactions on Evolutionary
Computation 11(4) pp. 521-540 (2007)
• [5] Polly Matzinger. Tolerance, danger, and the extended family.
Annual Review of Immunology, 12:991–1045, April 1994
• [6] E. Hart and J. Timmis. Application Areas of AIS: The Past, Present
and the Future. Journal of Applied Soft Computing 8(1). pp.191-
201. 2008

Overview of Methods for
Engineering Autonomic Self-Aware
Systems
Evolutionary Algorithms
Designed by Gusz Eiben

The Main EC Metaphor
EVOLUTION
Environment
Individual
Fitness
PROBLEM SOLVING
Problem
Candidate Solution
Quality
Quality  chance for seeding new solutionsCV
Fitness  chances for survival and reproduction
Fitness in nature: observed, 2ndary, EC: primary

General scheme of EAs
Population
Parents
Parent selection
Survivor selection
Offspring
Recombination
(crossover)
Mutation
Intialization
Termination

EAs as problem solvers
• EAs fall into the category of “generate and test” algorithms
• They are stochastic, population-based algorithms
• Variation operators (recombination and mutation) create the
necessary diversity and thereby facilitate novelty
• Selection reduces diversity and acts as a force pushing quality
• Traditionally seen as optimizers (also in data mining
applications)
• Lately as the engine behind adaptation

Two pillars of evolution
There are two competing forces
 Increasing population
diversity by genetic
operators
 mutation
 recombination
Push towards novelty
 Decreasing population
diversity by selection
 of parents
 of survivors
Push towards quality

Common model of evolutionary
processes
• Population of individuals
• Individuals have a fitness
• Variation operators: crossover, mutation
• Selection towards higher fitness
– “survival of the fittest” and
– “mating of the fittest”
Neo Darwinism:
Evolutionary progress towards higher life forms
=
Optimization according to some fitness-criterion
(optimization on a fitness landscape)

Main EA components
• Representation
• Population
• Selection (parent selection, survivor selection)
• Variation (mutation, recombination)
• Initialisation
• Termination condition

Representation
• Role: provides code for candidate solutions that can be
manipulated by variation operators
• Leads to two levels of existence
– phenotype: object in original problem context, the outside
– genotype: code to denote that object, the inside (chromosome,
“digital DNA”)
• Implies two mappings:
– Encoding : phenotype=> genotype (not necessarily one to one)
– Decoding : genotype=> phenotype (must be one to one)
• Chromosomes contain genes, which are in (usually fixed)
positions called loci (sing. locus) and have a value (allele)

Genotype spacePhenotype space
Encoding
(representation)
Decoding
(inverse representation)
B 0 c 0 1 c d
G 0 c 0 1 c d
R 0 c 0 1 c d
Representation 2
In order to find the global optimum, every feasible solution
must be represented in genotype space

Evaluation (fitness) function
• Role:
– Represents the task to solve, the requirements to adapt to (can be
seen as “the environment”)
– enables selection (provides basis for comparison)
– e.g., some phenotypic traits are advantageous, desirable, e.g. big ears
cool better, hese traits are rewarded by more offspring that will
expectedly carry the same trait
• A.k.a. quality function or objective function
• Assigns a single real-valued fitness to each phenotype which
forms the basis for selection
– So the more discrimination (different values) the better
• Typically we talk about fitness being maximised
– Some problems may be best posed as minimisation problems, but
conversion is trivial

Population
• Role: holds the candidate solutions of the
problem as individuals (genotypes)
• Formally, a population is a multiset of
individuals, i.e. repetitions are possible
• Population is the basic unit of evolution, i.e.,
the population is evolving, not the individuals
• Selection operators act on population level
• Variation operators act on individual level

Population 2
• Some sophisticated EAs also assert a spatial
structure on the population e.g., a grid
• Selection operators usually take whole
population into account i.e., reproductive
probabilities are relative to current generation
• Diversity of a population refers to the number
of different fitnesses / phenotypes /
genotypes present (note: not the same thing)

Selection
Role:
• Identifies individuals
– to become parents
– to survive
• Pushes population towards higher fitness
• Usually probabilistic
– high quality solutions more likely to be selected than low quality
– but not guaranteed
– even worst in current population usually has non-zero
probability of being selected
• This stochastic nature can aid escape from local optima

Example: roulette wheel selection
fitness(A) = 3
fitness(B) = 1
fitness(C) = 2
A C
1/6 = 17%
3/6 = 50%
B
2/6 = 33%
Selection mechanism example
In principle, any selection mechanism can be used for
parent selection as well as for survivor selection

Survivor selection
• A.k.a. replacement
• Most EAs use fixed population size so need a way of going
from (parents + offspring) to next generation
• Often deterministic (while parent selection is usually
stochastic)
– Fitness based : e.g., rank parents+offspring and take best
– Age based: make as many offspring as parents and delete all
parents
• Sometimes a combination of stochastic and deterministic
(elitism)

Variation operators
• Role: to generate new candidate solutions
• Usually divided into two types according to their arity
(number of inputs):
– Arity 1 : mutation operators
– Arity >1 : Recombination operators
– Arity = 2 typically called crossover
– Arity > 2 is formally possible, seldomly used in EC
• There has been much debate about relative importance of
recombination and mutation
– Nowadays most EAs use both
– Variation operators must match the given representation

Mutation
• Role: causes small, random variance
• Acts on one genotype and delivers another
• Element of randomness is essential and differentiates it from
other unary heuristic operators
• Importance ascribed depends on representation and
historical dialect:
– Binary GAs – background operator responsible for preserving and
introducing diversity
– EP for FSM’s/ continuous variables – only search operator
– GP – hardly used
• May guarantee connectedness of search space and hence
convergence proofs

before
1 1 1 0 1 1 1
after
1 1 1 1 1 1 1
Mutation example

Recombination
• Role: merges information from parents into offspring
• Choice of what information to merge is stochastic
• Most offspring may be worse, or the same as the parents
• Hope is that some are better by combining elements of
genotypes that lead to good traits
• Principle has been used for millennia by breeders of plants
and livestock

1 1 1 1 1 1 1 0 0 0 0 0 0 0
Parents
cut cut
Offspring
Recombination example
1 1 1 0 0 0 0 0 0 0 1 1 1 1

Initialisation / Termination
• Initialisation usually done at random,
– Need to ensure even spread and mixture of possible allele values
– Can include existing solutions, or use problem-specific heuristics, to
“seed” the population
• Termination condition checked every generation
– Reaching some (known/hoped for) fitness
– Reaching some maximum allowed number of generations
– Reaching some minimum level of diversity
– Reaching some specified number of generations without fitness
improvement

What are the different types of EAs
• Historically different flavours of EAs have been associated
with different data types to represent solutions
– Binary strings : Genetic Algorithms
– Real-valued vectors : Evolution Strategies
– Finite state Machines: Evolutionary Programming
– LISP trees: Genetic Programming
• These differences are largely irrelevant, best strategy
– choose representation to suit problem
– choose variation operators to suit representation
• Selection operators only use fitness and so are independent
of representation

Typical behaviour of an EA
Stages in optimising on a 1-dimensional fitness landscape
Early stage:
quasi-random population distribution
Mid-stage:
population arranged around/on hills
Late stage:
population concentrated on high hills

Typical run: progression of fitness
Typical run of an EA shows so-called “anytime behavior”
Bestfitnessinpopulation
Time (number of generations)

Evolutionary Algorithms in context
• There are many views on the use of EAs as robust
problem solving tools
• For most problems a problem-specific tool may:
– perform better than a generic search algorithm on most
instances,
– have limited utility,
– not do well on all instances
• Evolutionary Algorithms:
– Provide rather stable performance over a range of problems and
instances
– Can be easily ported from one problem to another one
– Can be hybridized with existing problem specific heuristics

Scale of “all” problems
Performanceofmethodsonproblems
Random search
Special, problem tailored method
Evolutionary algorithm
EAs as problem solvers: Goldberg view
(1989)

Summary
Evolutionary Algorithms:
• Are population-based, stochastic search methods
• Can be used for optimization and as “adaptive engine”
• Can be hybridized with problem specific heuristics
• Well suited for parallel execution
• Can be easily ported to new problems
An EA is the second best solver for any problem

Further reading
• A.E. Eiben, Evolutionary computing: the most
powerful problem solver in the universe?,
Dutch Mathematical Archive (Nederlands
Archief voor Wiskunde), vol. 5/3, nr. 2, 126-
131, 2002
• Eiben-Smith book

Academic Course: 05 Overview of Methods for Engineering Autonomic Self-Aware Systems

Recomendados

Recomendados

Mais conteúdo relacionado

Mais de FET AWARE project - Self Awareness in Autonomic Systems

Mais de FET AWARE project - Self Awareness in Autonomic Systems (20)

Último

Último (20)

Academic Course: 05 Overview of Methods for Engineering Autonomic Self-Aware Systems