Download OFAI - MindRACES

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Österreichisches Forschungsinstitut für
Artificial Intelligence
OFAI
AIS - Framework and Scenario
Lisbon September 2005
Overview
Österreichisches Forschungsinstitut für
Artificial Intelligence
• What have we presented so far?
• Immune Engineering –
a brief introduction
• AISA –
Artificial Immune System for
Anticipation
• The game room-scenario
2
Overview
Österreichisches Forschungsinstitut für
Artificial Intelligence
• What have we presented so far?
• Immune Engineering –
a brief introduction
• AISA –
Artificial Immune System for
Anticipation
• The game room-scenario
3
Level 1
Österreichisches Forschungsinstitut für
Artificial Intelligence
What have we presented so far?
Gibson 1966:
„Our perception of the world is expressed in terms of our interactions with it“
4
Level 2
Österreichisches Forschungsinstitut für
Artificial Intelligence
What have we presented so far?
5
Level 3
Österreichisches Forschungsinstitut für
Artificial Intelligence
What have we presented so far?
Level 2
6
Overview
Österreichisches Forschungsinstitut für
Artificial Intelligence
• What have we presented so far?
• Immune Engineering –
a brief introduction
• AISA –
Artificial Immune System for
Anticipation
• The game room-scenario
7
Österreichisches Forschungsinstitut für
Artificial Intelligence
Immune Engineering –
a brief introduction
• What is the Immune System?
• Anatomy of the Immune
System
• Immune Engineering
8
Österreichisches Forschungsinstitut für
Artificial Intelligence
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)
9
Österreichisches Forschungsinstitut für
Artificial Intelligence
What is the Immune System
• “The immune system is a cognitive
system whose primary role is to
provide body maintenance”
(Cohen)
10
What is the Immune System –
From a computational view
Systems that are:
Österreichisches Forschungsinstitut für
Artificial Intelligence
Computational Properties
•
•
•
•
•
•
•
•
•
Unique to individuals
Distributed
Imperfect Detection
Anomaly Detection
Learning/Adaption
Memory
Feature Extraction
Diverse
..and more
•
•
•
•
Robust
Scalable
Flexible
Exhibit graceful
degradation
• Homeostatic
11
Anatomy of the Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• The innate Immune System
• The adaptive Immune System
12
Österreichisches Forschungsinstitut für
Artificial Intelligence
The innate Immune System
• The part of the
immune system with
which we are born
• Doesn’t change or
adapt
• Complement system
• Macrophages
• Cytokines, natural
killer cells
13
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System
• Adapts to recognise
specific pathogens
• Retains a memory which
speeds up future
responses
• Consists primarily of
lymphocytes which
circulate around the body
• Distributed detection
system without any
central control
14
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System
• B lymphocytes
15
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• T lymphocytes
• helper cells
• killer cells
• suppressor cells
T-cell
antigen
16
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
epitope
• A detection event occurs
when the receptor of a
lymphocyte lymphocyte binds to an
epitope on a pathogen
receptor
• Strength of the bond is
termed the affinity
• Each lymphocyte can
have multiple, identical
receptors
Low affinity
structurally similar – high affinity
17
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System Generating Receptor Diversity
18
Österreichisches Forschungsinstitut für
Artificial Intelligence
Affinity Maturation – A Darwinian
process of variation and selection
Initial B-cell
population
Clonal selection
and
hypermutation
Adapted B-Cell
population
19
The adaptive Immune System Affinity Maturation
Österreichisches Forschungsinstitut für
Artificial Intelligence
B-Lymphocyte
Antigen
Clonal Selection
20
The adaptive Immune System Affinity Maturation
Österreichisches Forschungsinstitut für
Artificial Intelligence
B-Lymphocyte
high affinity
Antigen
Clonal Selection
21
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System Affinity Maturation
high affinity
M
Plasma cell
Memory cell
Clonal Selection
22
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System –
Jerne‘s Idiotypic Network
• The cells in the immune system can recognise
each other as well as pathogens!
• An antibody has a paratope and an idiotope
• Paratope-idiotope interactions leads to networks
of interconnected antibodies
23
Österreichisches Forschungsinstitut für
Artificial Intelligence
Jerne‘s Idiotypic Network
24
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System –
Jerne’s Idiotypic Network
• There is no essential difference between the
“recognized” and the “recognizer,”
• Any given antibody might serve either, or both,
functions.
25
Österreichisches Forschungsinstitut für
Artificial Intelligence
The adaptive Immune System
• Idiotypic Networks
26
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Idiotypic Networkdynamic
Idoptope-type 1
Paratope-type 1
Antigen
(epitop)
Recognising Set
27
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Idiotypic Networkdynamic
Idoptope-type 1
Idoptope-type 2
Paratope-type 2
Paratope-type 1
Internal Image
Antigen
(epitop)
Recognising Set
Paratopes p1 are recognised
by the idiotopes p2
28
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Idiotypic Networkdynamic
Idoptope-type 1
Paratope-type 1
Idoptope-type 3
Antigen
(epitop)
Paratope-type 3
Recognising Set
Idiotopes i1 are recognised by
the paratopes p3
Anti-idiotypic set
29
The adaptive Immune System
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Idiotypic Networkdynamic
Idoptope-type 3
Paratope-type 3
Idoptope-type 1
Paratope-type x
Anti-idiotypic set
paratopes p3 recognise
another set of idiotopes i1
Unspecific Set
30
Österreichisches Forschungsinstitut für
Artificial Intelligence
Immune Engineering
• AIS framework proposed by
De Castro and Timmis
•
31
Österreichisches Forschungsinstitut für
Artificial Intelligence
Immune Engineering
• Schematic
representation of an
AIS architecture
32
Österreichisches Forschungsinstitut für
Artificial Intelligence
Immune Engineering
• T-Cells - General Suppression
Framework
• Introduced by Luh and Liu in 2004
33
Overview
Österreichisches Forschungsinstitut für
Artificial Intelligence
• What have we presented so far?
• Immune Engineering –
a brief introduction
• AISA –
Artificial Immune System for
Anticipation
• The game room-scenario
34
Im m u n e L a ye r
c lo n a l s e le c tio n ,
m u ta tio n ,
n e tw o rk s d y n a m ic s ,
n e g a tiv e a n d p o s itiv e s e le c tio n ,
M
T-C e lls :
G e n e ra l S u p p re s s io n
AISA
A ffin ity L a ye r
Österreichisches Forschungsinstitut für
Artificial Intelligence
E u clid e a n ,
H a m m in g ,
M a n h a tta n ,
S im ila r,
C o m p le m e n ta r y
R e p r e se n ta tio n L a ye r
A n tig e n r e p r e se n ts th e
co m p le te d a ta u n ive r se
M u ltip le A n tig e n s fo r
e a ch situ a tio n
B - C e lls
S e n so r a n d D a ta - M e m o r y P o o l / D a ta U n ive r se
In fr a r e d
U ltr a so n ic
V id e o
V ir tu a l S e n so r s
F ilte r n
F ilte r 2
F ilte r 1
35
Österreichisches Forschungsinstitut für
Artificial Intelligence
AISA
• Data Universe
36
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Representation Layer
A n tig e n r e p r e se n ts th e
co m p le te d a ta u n ive r se
M u ltip le A n tig e n s fo r
e a ch situ a tio n
B - C e lls
37
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Representation Layer - Antibody
Robot Action
Command
C
A
E
partial description of
the sensory effects E
of doing Action A
Description C of
sensory information
38
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Representation Layer
•
Krautmacher and Digler 2004 –
AIS based navigation in a Rescue Scenario
•
Each situation is represented by
• many different antigens and
• antigen concentration (object distance)
39
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Representation Layer
•
Krautmacher and Digler 2004 –
AIS based navigation in a Rescue Scenario
40
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Affinity Layer
• Computationally, the degree of interaction
of an antibody-antigen or antibodyantibody can be evaluated by a distance or
affinity measure
• The choice of affinity measure is crucial:
• It alters the shape-space topology
• It will introduce an inductive bias into the
algorithm
• It needs to take into account the dataset used and the problem you are trying
to solve
41
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Affinity Layer
•
•
•
•
•
Euclidean
Hamming
Manhattan
Similar
Complementary
42
AISA
Österreichisches Forschungsinstitut für
Artificial Intelligence
• Immune Layer
•
•
•
•
•
•
•
Clonal Selection
Mutation
Bone Marrow Models
Immune Network Models
Negative Selection
Positive Selection
B-Cells - General Suppression
43
Overview
Österreichisches Forschungsinstitut für
Artificial Intelligence
• What have we presented so far?
• Immune Engineering –
a brief introduction
• AISA –
Artificial Immune System for
Anticipation
• The game room-scenario
44
Österreichisches Forschungsinstitut für
Artificial Intelligence
The game room-scenario
• Test bed - objects
45
Österreichisches Forschungsinstitut für
Artificial Intelligence
The game room-scenario
• Test bed - objects
46
Österreichisches Forschungsinstitut für
Artificial Intelligence
The game room-scenario
• Test bed - objects
47
Österreichisches Forschungsinstitut für
Artificial Intelligence
The game room-scenario
• Test bed - confinement
48
Österreichisches Forschungsinstitut für
Artificial Intelligence
The game room-scenario
• The Robots: AIBO
49
Österreichisches Forschungsinstitut für
Artificial Intelligence
The room-scenario
• The robots: KURT3D
50
Österreichisches Forschungsinstitut für
Artificial Intelligence
Prediction Fractal Machines
•
•
•
•
•
•
2nd approach
Proposed by Georg Dorffner
coordinated with IDSIA
2nd MindRaces year
More to follow …
Details see D4.1
51
Österreichisches Forschungsinstitut für
Artificial Intelligence
Discussion, Ideas, Questions,…