Artificial Immune Systems - Semantic Scholar
Artificial Immune Systems
Jon Timmis [email protected]
What is the Immune System ? a complex system of cellular and molecular components having the primary function of distinguishing self from not self and defense against foreign organisms or substances (Dorland's Illustrated Medical Dictionary) The immune system is a cognitive system whose primary role is to provide body maintenance (Cohen)
Immune system was evolutionary selected as a consequence of its first and primordial function to provide an ideal inter-cellular communication pathway (Stewart)
From a computational perspective: Computational Properties • • • • • • • • •
Unique to individuals Distributed Imperfect Detection Anomaly Detection Learning/Adaptation Memory Feature Extraction Diverse ..and more
Systems that are: • • • • •
Robust Scalable Flexible Exhibit graceful degradation Homeostatic
Reaction and memory
Immune memory is maintained via a network (this is a theory only)
Clonal expansion of cells in response to pathogenic exposure
Immune Learning
A Framework for AIS Solution Algorithms
Shape-Space Binary
AIS
Affinity
Integer Real-valued
Representation
Symbolic
Application [De Castro and Timmis, 2002]
A Framework for AIS Solution Algorithms AIS
Affinity
Euclidean
Representation
Hamming
Application
Manhattan
A Framework for AIS Solution Algorithms AIS
Affinity Representation
Application
Clonal Selection Negative Selection Immune Network Dendritic Cells T Cell Signalling
Immune Algorithms V Genetic Algorithms • Assume a similar representation you have seen for GAs • Fitness function concept is the same, except terminology is “affinity function” • Use of distance metrics is common • Mutation can be considered similar in terms of mechanisms, but when and how often it is done is very different • Selection mechanisms vary • Population sizes are dynamic and elements can interact
Clonal Selection Algorithms
Clonal Selection –CLONALG •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonalg •Initialisation •Antigenic
presentation
Create a random population of individuals (P)
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonalg •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
For each antigenic pattern in the data-set S do:
until stopping criteria
The Algorithms Layer
Conalg •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
Present it to the population P and determine its affinity with each element of the population
until stopping criteria
The Algorithms Layer
Clonal Selec*on nSelect
•Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
n highest affinity elements of P nGenerate clones proportional to their affinity with the antigen (higher affinity=more clones)
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
Mutate each clone High affinity=low mutation rate and viceversa Add mutated individuals to population P Reselect best individual to be kept as memory m of the antigen presented
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
Replace a number r of individuals with low affinity with randomly generated new ones
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Analysis • These can be analysed using Markov chains (see lecture 3 on gene*c algorithms) • Have similar proper*es to GAs, typically fewer fitness evalua*ons (but experiments have been selec*ve) • Convergence wise, they are the same • Muta*on opera*ons is where there is significant difference and this can affect performance and theore*cal analysis
Clonal selection variants
AIRS Algorithm • • • • • •
Static learning system Artificial Immune Recognition System (AIRS) Clonal selection based Uses the concepts of ARB’s (Artificial Recognition Balls) Resource based competition for ARB survival One-shot learning system
Memory Cell Identification A
Memory Cell Pool
MCmatch Found A
1
Memory Cell Pool MCmatch
ARB Generation A
1
Memory Cell Pool MCmatch Mutated Offspring
2
ARB Pool
Exposure of ARBs to Antigen 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
ARB Pool
Development of a Candidate Memory Cell 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
ARB Pool
Comparison of MCcandidate and MCmatch 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
4 2
A
MC candidate
ARB Pool
Memory Cell Introduction 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
4
5
A
MCcandidate
ARB Pool
Dynamic Learning AISEC: Email Classification
Continuous Learning Used when you want to classify changes over time (the notion of what is in a class) •
Levels of what you are interested in may change over time or the context of where you are working or what you are doing •
•
Web content mining is a perfect testbed for these ideas
• This study looked at email classification of interesting v un-interesting
AISEC •
• • • • • •
•
Supervised classification algorithm E-mail classified as interesting and uninteresting Uses constant feedback from user Capable of continuous adaptation This tracks concept drift and can also handle concept shift Representation: Subject, Sender and Return address Affinity measure: Proportion of words found in one cell compared to another (very naive) Mutation: Keep a library of words to use
The algorithm - classification System is initialised with known uninteresting e-mail Memory cells
Naive cells
E-mail presented for classification. Classified as uninteresting as it stimulates close cells
The algorithm – correct classification Highly stimulated cell reproduces n-times. Less stimulated cell produces fewer clones but with higher mutation rate Stimulation Region Classification Region
Cell with highest affinity is known to be useful therefore rewarded by becoming memory cell.
The algorithm cont… Incorrect classification Any cell responsible for incorrect classification is removed (memory or otherwise)
Cell removal Aged naïve cells deleted. Memory cells placed in already covered areas also deleted.
Summary • Immune inspired algorithms – focused on clonal selec*on • Similar in nature to gene*c algorithms • Other immune algorithms are very different – Immune network – T Cell signalling algorithms – Dendri*c cell algorithms
Jon Timmis [email protected]
What is the Immune System ? a complex system of cellular and molecular components having the primary function of distinguishing self from not self and defense against foreign organisms or substances (Dorland's Illustrated Medical Dictionary) The immune system is a cognitive system whose primary role is to provide body maintenance (Cohen)
Immune system was evolutionary selected as a consequence of its first and primordial function to provide an ideal inter-cellular communication pathway (Stewart)
From a computational perspective: Computational Properties • • • • • • • • •
Unique to individuals Distributed Imperfect Detection Anomaly Detection Learning/Adaptation Memory Feature Extraction Diverse ..and more
Systems that are: • • • • •
Robust Scalable Flexible Exhibit graceful degradation Homeostatic
Reaction and memory
Immune memory is maintained via a network (this is a theory only)
Clonal expansion of cells in response to pathogenic exposure
Immune Learning
A Framework for AIS Solution Algorithms
Shape-Space Binary
AIS
Affinity
Integer Real-valued
Representation
Symbolic
Application [De Castro and Timmis, 2002]
A Framework for AIS Solution Algorithms AIS
Affinity
Euclidean
Representation
Hamming
Application
Manhattan
A Framework for AIS Solution Algorithms AIS
Affinity Representation
Application
Clonal Selection Negative Selection Immune Network Dendritic Cells T Cell Signalling
Immune Algorithms V Genetic Algorithms • Assume a similar representation you have seen for GAs • Fitness function concept is the same, except terminology is “affinity function” • Use of distance metrics is common • Mutation can be considered similar in terms of mechanisms, but when and how often it is done is very different • Selection mechanisms vary • Population sizes are dynamic and elements can interact
Clonal Selection Algorithms
Clonal Selection –CLONALG •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonalg •Initialisation •Antigenic
presentation
Create a random population of individuals (P)
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonalg •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
For each antigenic pattern in the data-set S do:
until stopping criteria
The Algorithms Layer
Conalg •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
Present it to the population P and determine its affinity with each element of the population
until stopping criteria
The Algorithms Layer
Clonal Selec*on nSelect
•Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
n highest affinity elements of P nGenerate clones proportional to their affinity with the antigen (higher affinity=more clones)
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
Mutate each clone High affinity=low mutation rate and viceversa Add mutated individuals to population P Reselect best individual to be kept as memory m of the antigen presented
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
Replace a number r of individuals with low affinity with randomly generated new ones
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Clonal Selec*on •Initialisation •Antigenic
presentation
–Affinity evaluation –Clonal selection and expansion –Affinity maturation –Metadynamics •Repeat
until stopping criteria
The Algorithms Layer
Analysis • These can be analysed using Markov chains (see lecture 3 on gene*c algorithms) • Have similar proper*es to GAs, typically fewer fitness evalua*ons (but experiments have been selec*ve) • Convergence wise, they are the same • Muta*on opera*ons is where there is significant difference and this can affect performance and theore*cal analysis
Clonal selection variants
AIRS Algorithm • • • • • •
Static learning system Artificial Immune Recognition System (AIRS) Clonal selection based Uses the concepts of ARB’s (Artificial Recognition Balls) Resource based competition for ARB survival One-shot learning system
Memory Cell Identification A
Memory Cell Pool
MCmatch Found A
1
Memory Cell Pool MCmatch
ARB Generation A
1
Memory Cell Pool MCmatch Mutated Offspring
2
ARB Pool
Exposure of ARBs to Antigen 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
ARB Pool
Development of a Candidate Memory Cell 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
ARB Pool
Comparison of MCcandidate and MCmatch 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
4 2
A
MC candidate
ARB Pool
Memory Cell Introduction 1
A
Memory Cell Pool MCmatch
3
Mutated Offspring
2
4
5
A
MCcandidate
ARB Pool
Dynamic Learning AISEC: Email Classification
Continuous Learning Used when you want to classify changes over time (the notion of what is in a class) •
Levels of what you are interested in may change over time or the context of where you are working or what you are doing •
•
Web content mining is a perfect testbed for these ideas
• This study looked at email classification of interesting v un-interesting
AISEC •
• • • • • •
•
Supervised classification algorithm E-mail classified as interesting and uninteresting Uses constant feedback from user Capable of continuous adaptation This tracks concept drift and can also handle concept shift Representation: Subject, Sender and Return address Affinity measure: Proportion of words found in one cell compared to another (very naive) Mutation: Keep a library of words to use
The algorithm - classification System is initialised with known uninteresting e-mail Memory cells
Naive cells
E-mail presented for classification. Classified as uninteresting as it stimulates close cells
The algorithm – correct classification Highly stimulated cell reproduces n-times. Less stimulated cell produces fewer clones but with higher mutation rate Stimulation Region Classification Region
Cell with highest affinity is known to be useful therefore rewarded by becoming memory cell.
The algorithm cont… Incorrect classification Any cell responsible for incorrect classification is removed (memory or otherwise)
Cell removal Aged naïve cells deleted. Memory cells placed in already covered areas also deleted.
Summary • Immune inspired algorithms – focused on clonal selec*on • Similar in nature to gene*c algorithms • Other immune algorithms are very different – Immune network – T Cell signalling algorithms – Dendri*c cell algorithms
