Upload
icac09
View
1.144
Download
0
Embed Size (px)
DESCRIPTION
Citation preview
1
Dr. Perambur S. Neelakanta, Ph.D., C. Eng., Fellow IEEProfessor
Department of Electrical EngineeringCollege of Engineering and Computer Science
Florida Atlantic UniversityBoca Raton, Florida 33431, USA
In search of a cyberspace … to launch biologically-inspired advanced computing strategies:
a digital ecology solution
Invited LectureInternational Conference on Advanced Computing (ICAC 2009),
August 7-8, 2009, Tiruchirappalli, Tamil Nadu, India
1
2
Simply, known as bio-inspired computing (or just bio-computing), BIC denotes…
“a field of study that loosely knits together subfields related to the topics of connectionism, social behavior and emergence.
It is often closely related to the field of artificial intelligence, as many of its pursuits can be linked to machine learning.
It relies heavily on the fields of biology, computer science and mathematics…”.
In nut-shell, BIC is the use of computers to model nature, and simultaneously the study of nature to improve the usage of computers. It is, therefore a major subset of natural computation.
Biologically-inspired computing (BIC)…?
2
3
In search of a cyberspace ……to launch biologically-inspired advanced computing strategies….
Whether the strategies of BIC comes within the purview of information technology (IT)-oriented considerations is still unclear and remains as an open-question.
This paper heuristically searches for a cyberspace wherein BIC efforts can be viewed cohesively in the broader sense of IT-paradigms.
Hence, attempted here is an exploration to cast comprehensively the universe of BIC in the domain of so-called…
“digital ecology” (DE)
Now what is “digital ecology”?
3
4
Now what is “digital ecology”?
Digital ecology (DE) is a neoteric terminology mostly applied to the evolution of social and civic ecosystem commensurate with modern IT perspectives Its usage in modern context includes the plethora of (i) entertainment media ecology, (ii) the entirety of computing ambient and (iii) the environment of communication networks. In each of this gamut, the transfer of information (or informatics) negotiates a sizable cardinality of stochastically interacting stochastically interacting subsets that structuralize a complex open-source network and computational environment.
4
5
In short, DE refers to an environment, which is:- openin visibly portraying the interactions involved;- loosely-coupled in mediating the open relationships
between species;- domain-clustered in creating a field of balanced
common interest;- demand-driven in conglomerating the species as
interest groups;- self-organizing in autonomous decision-making; and,- agent-based in rendering an ambient of synergism
between human and machines where each agent participates proactively in the computational endeavors as well as in the information transfers akin to the species of biological ecosystem.
Now what is “digital ecology”? … Continued
5
6
Digital ecology enables a unified presentation of computational tools and algorithmic endeavors modern and advanced computing schemes) in an IT-specific domain. So attempted here in an ambient of BIC efforts towards…
… constructing a DE platform to support BIC concepts
As an illustrative example, the strategy of artificial neural networks (ANN) mapped in terms of relevant ontological norms of digital ecology is presented.
“Digital ecology” …a cyberspace to launch biologically-inspired advanced computing strategies
6
7
BIC bears the perspectives of cybernetics in the computational efforts involving …
� simulated annealing � artificial neural networks � genetic algorithms � DNA and molecular computing � biological ecology etc. Thus, the field of BIC is highly multidisciplinary, attracting a host of disciplines…- …computer science, molecular biology, genetics,
engineering, mathematics, physics, chemistry and others.
Biologically-inspired computing (BIC)…More
7
8
� DNA computation� nanofabrication� storage devices� sensing� healthcare � basic scientific research– for example …
…providing biologists with an IT-oriented paradigm to look at how cells “compute” or process the information
…helping computer scientists and engineers to construct algorithms based on natural systems, such as evolutionary and genetic algorithms
Biologically-inspired computing (BIC)……potential applications in:
8
9
BIC… its scope � Enabling new themes of computing technologies and fresh areas of computer science using biology or biological processes as metaphor/inspiration� Expanding information science concepts and tools to explore biology from a different theoretical perspective.
BIC as such, however, does not include in its scope the framework of, (i) the general use of computers; (ii) the strategies of computational analyses, and/or (iii) data management in biology - for example, bioinformatics or computational biology.
Biologically-inspired computing (BIC)…9
10
Genetic algorithms (GAs)↔ Follows natural evolution with the rules of selection, recombination, reproduction, mutation and more recently transposition. Such simple rules of evolution in complex organisms are observed and adopted in GAs constituting BIC approach.Artificial Intelligence (AI) ↔↔↔↔ Traditional AI is the intelligence of machines towards the design of intelligent agents. The way in which BIC differs from traditional AI is in how it takes a more evolutionary approach to learning, as opposed to what could be described as 'creationist' methods used in traditional AI. In this perspective AI inclines towards BIC.
Biologically-inspired computing (BIC)…… BIC and its cousins: Areas of emphasis
10
11
� Biodegradability prediction ↔ Accurate sequence details and genetic information vis-à-vis biodegradation are essential for assessing molecular basis of enzyme specificity, their catalytic mechanism, the evolutionary origin of related metabolism and proliferation of such activities in the environment.
(Although some basic formalization toward useful tools as a predictor of chemical/biodegradability is feasible, the absence of information at the sequence level of proteins etc. are imminently required for systematic studies of biodegradation. This is facilitated via biocomputing).
BIC and its cousins: Areas of emphasis… continued
11
12
Cellular automata ↔ Cellular automaton is a discrete model of a regular grid of cells, each in one of a finite number of states.
Relevant evolutionary computation programs with cellular arrays in decentralized platforms (where the information processing occurs in the form of global and local pattern dynamics) lead to emergent computation (expressed in terms of GAs) and adopted to evolve patterns in cellular automata in the perspectives of BIC.
BIC and its cousins: Areas of emphasis… continued
12
13
BIC and its cousins: Areas of emphasis… continued
Emergent systems↔ The way complex systems and patterns arise out of a multiplicity of relatively simple interactions as in biological systems is specified by “emergence”.
It has been the holy grail of BIC. Emergence is something like a macro phenomenon that appears as a by-product of a (generally but not always large) collection of micro phenomena.
13
14
BIC and its cousins: Areas of emphasis… continued
Neural networks ↔ Biological neural networks are made up of real biological neurons that are connected or functionally related in the peripheral nervous system or the central nervous system.
Artificial neural networks (ANNs) are composed of simulated neuron units made “in the image of real neurons”. By interconnecting “artificial neurons” – a programming strategy is set up that constructs a massively parallel connectivity, mimicing the biological neurons.
ANN with its interconnected structure of artificial neuron uses a paradigm of mathematical or computational model for information processing based on a connectionist approach to computation adaptively to changes in external or internal information via biological-inspiration.
14
15
�Artificial life ↔ Commonly known as Alife or alife, it depicts a field of study and an associated art form which examine systems related to life, its processes, and its evolution through simulations using computer models, robotics, and biochemistry.
There are three major versions of alife, based on their approaches: soft- from software; hard- from hardware ; and wet- from biochemistry. Artificial life imitates traditional biology in recreating biological phenomena. Essentially, the term "artificial life" is often used to specifically refer to soft alife.
BIC and its cousins: Areas of emphasis… continued
15
16
�Artificial immune systems (AIS) ↔ Abstracting and mapping the structure and function of an immune system to a computational set of frameworks so as to investigate the application of such systems towards solving computational problems with the aid of mathematics, engineering, and information technology.
AIS is a sub-field of computational intelligence, BIC, and natural computation, with a focus on machine learning. It can be said to belong the broader field of AI. Further, AIS are adaptive systems, inspired by theoretical immunology and observed immune functions, principles and models, which are applied to problem solving.
BIC and its cousins: Areas of emphasis… continued
16
17
Rendering (computer graphics) ↔ a process of generating an image from a model (description of 3D objects in a strictly defined language or data structure) using computer programs.
It contains features of geometry, viewpoint, texture, lighting, and shading information in digital image or a raster graphics image format.
The term rendering in computing context is an analogy of an "artist's rendering" of a scene. (In biological context, rendering simply refers to patterning and rendering of animal skins, bird feathers, mollusk shells and bacterial colonies)
BIC and its cousins: Areas of emphasis… continued
17
18
Lindenmeyer systems↔ Computing self-organization in the context of environmentally sensitive growth and/or development modeling behavior and visualization of cells of plants/plant structures: - Mathematical, spatial models that treat plant geometry as a continuum or as discrete components in space. - Developmental models that describe form as a result of growth in terms of growth influencing variables - Simulations produce numerical output, which can be complemented by rendered images and animations for the purpose of easy comprehension
BIC and its cousins: Areas of emphasis… continued
18
19
Communication networks and protocols↔ Analogy between viral dynamics in humans and in computers is useful in assessing infectious disease epidemiology on human social networks versus communication in wireless networks
Epidemiology as a metaphor may hold insights into communication networks.
New paradigms of mathematics and methodologies sought towards linking epidemiology and the spread of disease are generalized biological-inspirations seen toward modeling modern communication systems.
BIC and its cousins: Areas of emphasis… continued
19
20
Membrane computers↔ The membrane computing is an effort to replicate organic structures of the brain and the intra-membrane molecular processes in the living cells onto silicone.
This is to create indeterminate outcome machines that are capable of learning through external stimuli.
Such membrane computers will be a interesting technology when it is finally developed, say in creating artificial brains and teaching machines… a dream sought in BIC.
BIC and its cousins: Areas of emphasis… continued
20
21
Excitable media ↔ An excitable medium is a nonlinear dynamical system that has the capacity to propagate a wave of some description, experiencing an elapsed time (refractory time). A forest is an example of an excitable medium: That is, when a wildfire burns through the forest, no fire can return to a burnt spot until the vegetation has gone through its refractory period and re-grown. BIC implications are related to…
Pathological activities in the heart and brain can also be modeled as excitable media.
In cellular automata the state of a particular cell in the next time step depends on the state of the cells around it--its neighbors-at the current time.
BIC and its cousins: Areas of emphasis… continued
21
22
Sensor networks ↔↔↔↔ Sensor networks are a sensing, computing and communication infrastructure that allows to instrument, observe, and respond to phenomena in the natural environment, and in the physical as well as cyber infrastructure.
Akin to biological systems that present remarkable adaptation, reliability, and robustness in various environments, even under hostility (in a distributed and self-organized way), they provide useful resources for designing the dynamical and adaptive routing schemes of wireless mobile sensor networks.
BIC and its cousins: Areas of emphasis… continued
22
23
DNA computing ↔↔↔↔ a computing strategy that uses interdisciplinary aspects of DNA, biochemistry and molecular biology, instead of the traditional silicon-based computer technologies. It is a molecular computing stategy similar to parallel computing and employs many different molecules of DNA to try many different information processing at once. Mostly, DNA computers are faster and smaller than any other computer built so far. However, unlike quantum computing, in DNA machines to solve extremely large EXPSPACE problems, the amount of DNA required is too large to be practical.
BIC and its cousins: Areas of emphasis… continued
23
24
Biologically-inspired computing will be “wonderful tools, (and) will eventually lead the way to a “molecular revolution,” which ultimately will have a very dramatic effect on the world”. As such biocomputing, in general has the potential to be a very powerful tool.
BIC shouldering the marvels of computationper seis not the traditional “computing with silicon-chips”, but in essence, it. It relies on information-science (technology?) and borrows the metaphors from biological sciences.
The query that lingers is whether the various avenues of BIC can be comprehended in a unified cyberspace. If so, how?
BIC: CAN IT BE COMPREHENDED IN A UNIFIED CYBERSPACE?
24
25
In modern perspective, in sheltering the BIC within the scope of IT-oriented considerations is still unclear and remains as an open-question. Suppose BIC-related computational tools and algorithmic endeavors are to be viewed in an IT-specific cyberspace. It is then necessary to seek a platform that permits a cohesive activity of a complex system where biological evolutionary principles are invoked in terms of interacting species having self-organizing features. Further overlaid thereon are feasible aspects of informatics and paradigms of computation.
BIC: CAN IT BE COMPREHENDED IN A UNIFIED CYBERSPACE? …Continued
25
26
Can the underlying abstract of a unified cyberspace of BIC be specified in the so-called digital ecology (DE) platform towards a compatible solution?
DE is “the medley of digital code and environmentalism” that prescribes information ecosystems constituted by information flows being processed through various mediating species across biological ecology. In this perspective, considering the intersecting aspects of a complex system and ecological prescriptions, models of BIC can be projected in the realm of digital ecosystem ontology.
BIC: CAN IT BE COMPREHENDED IN A UNIFIED CYBERSPACE? …Continued
26
27
Digital ecosystems have been conceived in “the image of” complex biological ecology expressed in terms of "digital environment" ontology and is populated by "digital species" that mediate massive information exchange.
Compared with natural ecosystems where species may follow adaptation to local conditions, in digital ecosystem, newdigital species continuously emerge and they help cleanse the ecosystem (for example supplanting older scheme of computation with an advanced one).
Digital ecosystems thus capture the essence of classical, complex ecological environment in nature, where organisms cohesively constitute a dynamic, self-organizing and interrelated complex ecosystem conserving and utilizing the environment of its resources.
BIC: CAN IT BE COMPREHENDED IN A UNIFIED CYBERSPACE? …Continued
27
28
… a possible suite for modeling the complex system profile of BIC is to apply DE considerations identified in terms of certain DE ontology nomenclature:{Species} ⇔⇔⇔⇔ {Domain, Task, Profit, Rule, Role,
Supplier, Requester, Available Service, Requested Service}
{Environment} ⇔⇔⇔⇔ {Technology, Service, (Species), Open-environment, Loosely-coupled environment, Demand-drivenenvironment, Domain-clusteredenvironment}
BIC: A COMPLEX SYSTEM THAT FOLLOWS A DIGITAL ECOSYSTEM ONTOLOGY...
28
29
{Species} ⇒⇒⇒⇒ {Domain, Task, Profit, Rule, Role, Supplier, Requester, Available Service, Requested Service}
{Environment} ⇒⇒⇒⇒ {Open, loosely-coupled, demand-driven; domain-clustered}
⇑⇑⇑⇑ ⇓⇓⇓⇓{Interacting neurons, layered ANN architecture, massively parallel computation, output/goal-realization, nonlinear processing of collective information, supervised learning; output validation via teacher value, input ambient, user (programmer), convergence of the output against learned pattern, testing an input set against learned pattern}
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN EXAMPLE… ANN
29
30
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN ANN EXAMPLE…continued
+–
Inputsyi = f(xi)
Inputlayer
Hiddenlayers
Ouputlayer
Weights Weights
TeacherInput
ΣΣΣΣOi
Oi = KΣΣΣΣzi
T i
εεεεi = (Oi, Ti)Error
Weight vectoradjustments
A neuronal
unit
zi
30
31
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN ANN EXAMPLE…continued
CALL: DEFINE_ ENVIRONMENT: ANNDEFINE_SPECIES: Neuronal units ⇒⇒⇒⇒
comesFrom - domain ANN architectureDEFINE_DOMAIN: ⇒⇒⇒⇒ common field for all species
DEFINE_TASK carriesOut goal-oriented tasksGoal : converged ANN outputDEFINE_PROFIT relatesTo task- computational advantageisDrivenBy species : neuronsDEFINE_RULE:-follows nonlinear norms regulating species collectivelyDEFINE_ROLE- role of interaction with otherSpecies (neurons) defineBy weight-modification,inter-play of input data at the hidden layer(s)
CALL: DEFINE_ SUPPLIERCALL: DEFINE_ REQUESTER
(A) Subfunction PeudoCode I on:DEFINE_(ANN-DE)_SPECIES& ENVIRONMENT –ONTOLOGY: Initialize
⇒⇒⇒⇒ FOR Complex ANN system: Neurons/neuronal units
31
32
Inputs : Training and prediction sets: DEFINE_DIGITAL_ECOSYSTEM: ANNDEFINE_ENVIRONMENT⇒⇒⇒⇒ architecture items of SPECIES
DEFINE_TECHNOLOGY of the Environment isSupportedBy INPUTS and Teacher values
Connectivity isProvidedBy SPECIESGOTO: SPECIESDEFINE_SERVICESError feedback –backpropagation etc.Weighting is rendered on SPECIES/Interconnected
DEFINE_ENVIRONMENT set:{open, demand-driven, agent-based, self-organizing, domain-clustered, loosely-coupled}- ANN architecture
(B) Subfunction Code II on: ENVIRONMENT ontology -
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN ANN EXAMPLE…continued
32
Initialize:
33
Computation of: ANN Output
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN ANN EXAMPLE…continued
Inputs to: { Species and Environment}:←DOMAIN data set {details on neurons, layers, logistic func tion,
momentum function, learning coefficient}←←←← ENVIRONMENTdata set {Training data set to visible neurons,
teacher values}←←←← TASK data set {Defining error, type of feedback etc.}← RULE data set {Stop criterion on iterations, tuning the
weighting coefficients}← ROLE data set {Adjusting the nonlinearity, momentum and
learning towards convergence}←←←← REQUESTERdata set {Input data to visible neurons, teacher se t}}←←←← SUPPLIER data set {ANN user}Compute I : Related subfunctions towards output O i (t)←←←← REQUESTERobservation at the output nodeCompute II : IF computed error is too high, ←←←← THEN do iteration←←←← OR ELSE, GOTO Compute IEND
33
34
DEFINE_ROLE
DEFINE_SUPPLIER-ANN user
DEFINE_AVAILABLE_SERVICE- ANN capability
Subfunction Code IIA on: SUPPLIER suite of SPECIES ontology DEFINE_ROLE
Convergence toward objective function
DEFINE_REQUESTER⇒⇒⇒⇒ ANN output
DEFINE_REQUESTED_SERVICE⇒ Convergence
towards thegoal sought
Subfunction Code IIB on: REQUESTER suite of SPECIES ontology
Subfunction Codes on: SUPPLIER and REQUESTER
BIC: SPELT IN THE ONTOLOGY OF DIGITAL ECOLOGY – AN ANN EXAMPLE…continued
34
35
REFERENCES
[1] N. Forbes, Biologically inspired computing, Computing in Science and Engineering, November/December 2000, vol. 2(6), 84-87
[2] H. Boley and E. Chang, “Digital ecosystem: Principles and semantics,” in 2007 Inaugural IEEE International Conference in Digital Ecosystems and Technologies (IEEE DEST 2007), 2007, 1-4244-047003/07.
[3] H. Dong, F. K. Hussain, and E. Chong, “Ontology-based digital ecosystem conceptual representation,” in Proceedings of the Third International Conference on Automatic and Autonomous Systems (ICAS’07), 2007, 0-7695-2859-5/07
[4] P. S. Neelakanta and R. C. Tourinho, Modeling an It-centric complex system via digital ecology concepts, Presented in Third IEEE International Conference on Digital Ecosystems and Technologies (IEEE-DEST 2009), Istanbul, Turkey, 31 May 2009 – 3 June 2009)
[5] G. W. Flake: The Computational Beauty of Nature, MIT Press. Boston, MA: 2000 [6] P.S. Neelakanta and D. De Groff, Neural Network Modeling: Statistical Mechanics and
Cybernetic Perspectives, CRC Press, Boca Raton, FL, 1994.[7] P.S. Neelakanta, “Dynamics of neural learning in the information theoretic plane,” Chapter
5, Information-Theoretic Aspects of Neural Networks(Editor: P.S. Neelakanta), CRC Press, Boca Raton, FL, 1999.
[8] L. M. Adleman, Computing with DNA, Scientific American, August 1998, 54-61
35A
36
In search of a cyberspace … to launch BIC…Conclusions
This study attempts to portray biologically-motivated computing considerations…
… in the framework of a complex digital ecosystem.
… the ANN is chosen as an example and characterized in the domain of interest.
… Relevant details on ANN describe the relational aspects of Speciesand Environment vis-à-visthe BIC in terms of the ontological details of [3].
THANK YOU !!!
35B