Scholarly article on topic 'Effective cognitive characteristic in robustness of decentralized system'

Effective cognitive characteristic in robustness of decentralized system Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{Agent / "Capra Cogitive Framwork" / "immune system" / lymphocyte / robustness}

Abstract of research paper on Materials engineering, author of scientific article — Touraj Banirostam, Mehdi N. Fesharaki

Abstract Decentralized systems can be defined as systems without a central element, and robustness of such system would depend on several parameters. Introducing Capra Cognitive Framework, the Immune System has been assessed as a decentralized system which enjoys a higher robustness regarding the mentioned framework. The Immune System's functionality has been considered and modelled based on Capra Cognitive Framework. Using Biological Agents, the proposed model has been simulated and, then, some impacts of its cognitive characteristics on the system's robustness have been evaluated.

Academic research paper on topic "Effective cognitive characteristic in robustness of decentralized system"

SciVerse ScienceDirect PfOCSCliCI

Social and Behavioral Sciences

Procedia - Social and Behavioral Sciences 32 (2012) 134 - 140

4th International Conference of Cognitive Science (ICCS 2011)

Effective cognitive characteristic in robustness of decentralized system

Touraj Banirostama'*, Mehdi N. Fesharakia

aComputer Departement, Sicence and Research Branch, Islamic Azad University, Pounak, Tehran, Iran

Abstract

Decentralized systems can be defined as systems without a central element, and robustness of such system would depend on several parameters. Introducing Capra Cognitive Framework, the Immune System has been assessed as a decentralized system which enjoys a higher robustness regarding the mentioned framework. The Immune System's functionality has been considered and modelled based on Capra Cognitive Framework. Using Biological Agents, the proposed model has been simulated and, then, some impacts of its cognitive characteristics on the system's robustness have been evaluated.

© 2011 Publishied by Elsevier Ltd. Selection and/or peer-review under responsibility of the 4th International Conference of Cognitive Science

Keywords: Agent; Capra Cogitive Framwork; immune system; lymphocyte; robustness

1. Introduction

Decentralized systems have been considered as complex systems which their robustness appears to be based on the relations between the elements of the system. These systems would have a high level of interactions and also high number of elements which could be found as simple or complicated (Bar-Yam, 1997). Interactions in a system would lead every constitutive element to get formed and evolved through the time (Capra, 1996). In another word, in such a system, evolution contains both individual and social aspects. In every agent, the individual evolution should be considered as a result of learning and also the interactions that would lead to the social evolution and these two parameters could cause the evolution of the whole system (Bourgine & Stewart, 2004).

Studying decentralized biological systems could be effective in finding out the parameters of a decentralized system's robustness. The Immune System (IS) has been considered as a decentralized cognitive living system with high robustness. As a framework for social living systems, Capra Cognitive Framework (CCF) (Capra, 2002) has considered these systems through four points of view. The first two viewpoints (dimensions) are Pattern and Structure and the third one is Process. The Pattern dimension has focused on relations between elements of a system, the physical embodiment of an organization has formed the Structure and a system's Structure evolves through its lifetime. These two dimensions have been combined together in the dimension of Process. Generalizing this framework to the social-dominant dimension of Concept would be added to the mentioned three ones. Concept

* Corresponding author. Tel.: +98-9123587515; fax: +98-2144869655 E-mail address: t_banirostam@srbiau.ac.ir

1877-0428 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the 4th International Conference of Cognitive Science doi:10.1016/j.sbspro.2012.01.022

dimension also can lead to the appearance of convergence and robustness of the social system. Within real social systems and in order to make CCF adapt to the existed context, Network (the relations between the components of a system) has been used for the Pattern. Agent has been mentioned as the Structure and the Process has included agent's methods of decision making. The Process is a function of Structure and Network, which means that all affairs in the Process dimension would be affected by the Network and Agents of system. Adapted Capra Cognitive Framework (ACCF) for simulating a decentralized social system has been illustrated in Figure 1. More details about CCF and ACCF have been presented in (Banirostam & Fesharaki, 2011a; Capra, 2002).

Procrw

(Behaviour Ocnemion)

Fig. 1. Adaptive Capra Cognitive Framework (ACCF)

To recognize the characteristics effective on decentralized system robustness, a system with the same characteristics should be studied and modeled (Miller & Page, 2007). As a social system, the Immune System can be considered as a decentralized, cognitive and complex system which enjoys self-management characteristic.

In the following sections, a model based on ACCF for the IS is introduced. After that, a simulation design based on proposed model in Netlogo environment will be described and, finally, the results of the simulation will be presented.

2. Modeling the Immune System

To model IS, its functionality should be precisely studied. The details of this functionality have been mentioned in (Medzhitov & Janeway, 2000; Koyama, Ishii, Coban, & Akira, 2008; Eisenbarth & Flavell, 2009) and also some related works and IS modeling with different approaches have been considered in (Banirostam & Fesharaki, 2011b). As an independent entity, agent can be used for IS modeling and simulation and for IS modeling with ACCF, characteristics of the four dimensions should be determined and applied in each agent. Regarding the four dimensions of ACCF, IS modeling has been considered in the following subsections and moreover, characteristics of each dimension have been introduced.

2.1. Process Dimension- Modeling the behavior rf lymphocyte (Antibody)

Decision-making and behavior generation of an entity have been represented by the Process dimension. As the main elements of IS, lymphocytes (antibodies) are to be autonomous. Studying general behavior of lymphocytes, it can be observed that behavior generation in any of antibodies follows figure 2.

Fig. 2. General process of behavior generation

Each lymphocyte would pass four steps of Monitoring, Analyse, Planning and Execute to generate a behavior and this is to be mentioned that the knowledge in their knowledge base would be utilized in the last three steps. Having a thorough looking on behavior generation of each lymphocyte, it can be found that this process has been following

the OODA control loop (McMichael & Jarrad, 2005). In this process, Monitoring has been considered as Observation, Analyse as Orientation, Planning as Decision-making and Execution as Action in OODA control loop. In addition, activities bellow would take place in each step:

• Monitoring: Lymphocytes move randomly and monitor their environment moment by moment. To face or not to face other elements has been controlled within this step.

• Analyse: Through the random move and after confronting another element, lymphocytes recognize self and non-self elements and to recognize non-self elements, they check the protein chain of that element with knowledge in their own knowledge base. In this step, two processes of recognition and matching occur.

A- Recognition of self and non-self elements: In their random walk, each lymphocyte would compare other elements' protein chain with self elements' ones and if they were equal, the element would be recognized as self and otherwise, it will be recognized as a non-self element.

B- Matching process: If an element was non-self, lymphocyte would compare that element's protein chain with recorded knowledge in its own memory. Matching protein chain of a non-self element (antigen) means that the antigen has been recognized by the lymphocyte. In such situation, the lymphocyte would become active and then the active lymphocyte will enter the Planning step, migrate to the lymph nodes and do autopoiesis.

If the lymphocyte was not able to confront the antigen, it would save the antigen's information and enter the Planning step. New information will be sent to the other lymphocytes through communication.

• Planning: Lymphocytes enter this step just when they face a non-self element. If the lymphocyte would not be able to confront the recognized antigen, difference between antigen's protein chain and what recorded in the lymphocyte's knowledge base would be calculated and if the difference was less than a determined amount, the lymphocyte could become active by means of adaptation and gain the ability to confront the antigen. This time, adaptation would take place in the planning step and the lymphocyte adds the new antigen's information to its knowledge base and then, it migrates to the lymph nodes. Adaptation in lymphocytes means they have ability to change some knowledge in their knowledge base by mutation.

• Execute: After migration, lymphocytes would begin to reproduce themselves and after that, they will be scattered in the environment to attack the antigens.

In the presented model in figure 2, learning has been existed in both steps of analyse and planning. Furthermore, the communications would happen after facing a self element and also making the decision of autopoiesis would take place in planning step. Within the step of execution, the final behavior would be an emergence behavior (Moffat, Smith, & Witty, 2006) which would be applied and the feedback of the environment to this behavior could result in structural evolution in lymphocytes. A lymphocyte's behavior generation adapted with OODA control loop and with more details has been illustrated in figure 3 and described more in (Banirostam & Fesharaki, 2011a).

Fig. 3. Behavior generation of a lymphocyte

2.2 Structure Dimension- Lymphocyte's Attributes and Characteristics

As the only physical dimension, Structure can represent the constitutive elements of a system. Structures which contain the capacities of being autonomous and self-management should be used for modelling and applying a decentralized system. The main characteristics of lymphocytes' structure which constituted the IS are: mobility, knowledge base (memory), adaptation, learning, situation awareness, autopoiesis, inheritance and evolution. These characteristics should be applied in simulation agent and have been explained bellow.

• Mobility: Lymphocytes' movement can be defined as random walk.

• Knowledge Base (memory): Each lymphocyte's knowledge base contains a determined number of antigens' codes. Lymphocytes are able to add new recognized antigen's code to their knowledge base.

• Adaptation: The Immune Response comes out of comparing antigen's chain of protein with the existed chains in lymphocyte's knowledge base. In this situation, the lymphocytes adapt themselves by means of mutation which leads the structure to change and make the lymphocyte become active in future confrontments with the antigens. The conclusion of all this process can be mentioned as learning.

• Learning: Learning process (which means lymphocyte's structural changes) would cause the addition of new recognized antigen's chain of protein (code) to the lymphocyte's knowledge base. As time goes on, this process can result in the evolution of lymphocyte and therefore make the lymphocytes show a better reaction in the next confrontments.

• Situation Awareness: Doing random walk in the environment, every lymphocyte would become aware of the existence of other elements within that environment. In lymphocytes, this awareness could lead to learning, communications and autopoiesis. Also among antigens, this awareness could result in learning and structural change in the form of changing the chain of protein.

• Autopoiesis: Every lymphocyte/antigen is able to produce an entity quite like itself during some determined intervals. Through this process, the parent entity transmits its whole memory to the child one. For lymphocytes, this process would happen when they recognize an antigen and migrate to the lymph nodes. But for antigens, autopoiesis occurs in determined periods of time.

• Inheritance: During the process of autopoiesis, every lymphocyte or antigen transmits its characteristics to its child. This action could make new lymphocytes get stimulated by the recognized antigens.

• Evolution: lymphocytes could learn the actions and reactions of other elements. The process of learning would cause structural changes within the lymphocytes and enhance their functionality in future situations. Moreover, this characteristic makes the lymphocyte evolve through the time and also keep this evolution on to the next generations.

2.3. Pattern Dimension- Communication Network

Relations between elements of a system should be considered in this dimension. In social systems, the meaning of Pattern has been communicating network and its conclusions. Communication network in IS has been consisted of blood and lymph and the existence of such network in a social system like IS could be effective on robustness and convergence of the system. Furthermore, the existence of characteristics which have been mentioned bellow would be mentioned as the results of communication network in a system:

• Shared Situation Awareness: This situation comes after the awareness of each lymphocyte about the existence of an antigen and recognizing that. Shared Situation Awareness is the final result of communications between lymphocytes.

• Community of Interest (CoI): It means the common sense to conquest the antigens and this is to be the result of situation awareness, shared situation awareness and commitment.

• Interpersonal trust: One of the substantial parameters in social system is interactions based on trust. The existence of trust leads to the rapid formation of situation awareness and shared situation awareness. Lymphocytes are fully trusted in each other. Though this characteristic results in faster convergence, but this phenomenon also can prevent some diseases like cancer from getting not recognized by the IS components.

• Evolution: Every lymphocyte would evolve by means of learning ability. Also another kind of evolution can appear in a system through communication network which means that lymphocytes reach Shared Situation Awareness through interaction and communication with each other and this process can probably lead the system to some Structural changes within the whole components of IS.

2.4. Concept Dimension

Concept dimension represents the main objectives of a social system and results in system's robustness and convergence. Macro policies of a system would be defined in this dimension. Also the characteristic of self-protection within IS should be assessed in this dimension.

• Self-protection: This characteristic has been defined as an element's ability to defend itself against non-self elements without interference of the others. Because of the perpendicularity of Concept dimension on the other three dimensions (Agent, Network and Process), the context of self-protection shows up in all these three dimensions as well. In addition to the self-protection characteristic, characteristics below should be noticed in this dimension.

• Commitment: This characteristic means being committed to do assigned affairs and is also effective on system's robustness and convergence. Commitment has been represented in the system's general policies.

• Trust: Trust can be considered as the acceptance of functionality and data of self elements. In the Network dimension, this context has been illustrated in the form of interactions based on trust among self elements.

Due to Capra Cognitive Framework, as the dimension of Concept is going to get formed, the other three dimensions with represented characteristics would also get formed in the direction of this dimension. In another word, if there were agents with mentioned characteristics and presented rationality approach and the communicating network between them had noticed characteristics, then with applying Concept dimension, a system can be kept in a robust situation. At last this should be mentioned that dimension of Concept should be applied in dominated algorithm on agents.

3. Proposed Model Simulation and Results

Up to now, IS has been modeled with different methods and simulated by Cellular Automata (CA), agent and Multi Agent Systems (MAS). Related works on modeling and simulating IS has been presented in (Banirostam & Fesharaki, 2011a, 2011b). IS has been made up of heterogeneous, mobile and asynchrony elements and has had an uncertain structure. Therefore, CA can not have a high efficiency on this simulation (Bandini, Manzoni, & Vizzari, 2002).

CA, MAS and Situated Cellular agent (SCA) for simulating the IS might be able to illustrate some contexts such as learning, communication and emergence but utilizing Biological Agents can show a better representation of the IS functionality. More details about Biological Agent have been presented in (Banirostam & Fesharaki, 2011a).

Modeling the IS functionality can lead to recognition of effective characteristics on its robustness. Also through simulating, the amount of each characteristic's impact on system's robustness can be determined. Netlogo environment (Macal & North, 2006) has been used to simulate the proposed model with Biological Agents and noticed characteristics in part 2. There are two groups of Biological Agents instead of lymphocyte and antigen in the simulated environment which follow the diagram in figure 2 for behavior generation. All of the lymphocytes have the same code but antigens' codes are different and random. Each lymphocyte agent possesses a random knowledge of antigens' codes and every agent does random walk and therefore, all lymphocytes follow the diagram shown in figure 3 for behavior generation through their lifetime.

Considering that every agent (both lymphocytes and antigens) has the ability of autopoiesis, robustness can be defined as the ability to destruct all rival agents. The maximum time of robustness for lymphocyte and antigen agents is the time which at least one agent of each group can stay alive and still have the ability of autopoiesis. Also convergence appears when lymphocytes conquest all antigens or vice versa.

When simulation has been run in different situations, this can be observed that adaptation increases the system robustness. When the adaptation ability grows, the system robustness will increase too. Results of adaptation in different situations have been illustrated in figure 4. In this figure, X axis shows the percent of lymphocyte agents which have been adapted and the Y axis shows robustness time.

A lymphocyte's memory before and after running has been illustrated in figure 5. It can be observed that before running, a certain lymphocyte just has the ability to confront the antigens with the codes of 4, 5 and 8, but because of learning ability, after running the lymphocyte is able to recognize antigens with the codes of 4, 5, 8, 3, 1, 9 and 6.

'.. ill,with Mutation

Fig. 4. Robustness of the IS in different mutation (adaptation) rate

Fig. 5. Memory of antibody before and after running

Running the simulation, this can be figured out that learning ability is one of the essential components in robustness and evolution of a social system. Therefore, increasing lymphocytes' learning rate makes their robustness increase. Results of simulation in different situations of learning rate have been shown in table 1.

Table1. The effect of different learning ability in robustness of system

Learning Ability Without learning 1 4 7 10

System Robustness 700 1510 2000 2250 2480

Through the time, learning and communication lead to the evolution of system and this evolution increase the system robustness. Two different situations have been illustrated in figure 6. In situation A, lymphocytes has faced a new element and learned its functionality and in this situation, the average duration of robustness is about 1500 time units. But the unknown element has been recognized by the lymphocytes in situation B and lymphocytes could resist against that element about 2900 time units. In this situation, evolution could be the result of the learning ability, or the communication with other lymphocytes. In figure 6, X-axis shows time and Y-axis represents the population of lymphocyte and antigen.

Fig 6. Effect of learning and evolution in robustness of the system

The system's functionality in two situations (with and without communications) can be observed in figure 7. Comparing these two situations, the impact of communications on system's robustness can be found out.

Fig 7. Impact of Communication on IS robustness 4. Conclusion

Decentralized systems are self-organized systems which can manage themselves without any central element. Presenting Capra Cognitive Framework can show that the IS has been assessed as a decentralized social system with high robustness. Considering this framework, a social system has been made up of four dimensions: Process, Agent, Network and Concept. Through these four dimensions, the characteristics of IS have been evaluated and regarding to this process, the functionality of this system has been modeled. Based on Adapted Capra Cognitive Framework, if agents with the mentioned characteristics and rationality approach have existed and the communicating network between them had the noted attitudes, then, a decentralized system can be kept robust when the dimension of Concept is applied. Using Biological Agents, the proposed model has been simulated and the impact of some parameters on system robustness has been evaluated.

Bandini, S., Manzoni, S., & Vizzari, G. (2002). Situated cellular agents: A model to simulate crowding dynamics. IEICE Transactions on Information and Systems: Special Issues on Cellular Automata, E87-D, 669-676.

Banirostam, T., & Fesharaki, M. N. (2011a). Immune system simulation with biological agent based on capra cognitive framework. Proceedings of the 13th International Conference of Computer Modelling and Simulation (UKSim), UK, 122-127.

Banirostam, T., & Fesharaki, M. N. (2011b). Effective parameters in convergence of autonomous distributed systems using with immune system approch. Paper presentd at the IEEE ISADS 2011, Tokyo.

Bar-Yam, Y. (1997). Dynamics of complex systems. Reading, MA: Addision-Wesley.

Bourgine, P., & Stewart, J. (2004). Autopoiesis and cognition. Artificial Life, 10, 327-345.

Capra, F. (1996). The web of life: A new scientific understanding of living systems. New York: Doubleday.

Capra, F. (2002). The hidden connections: Integrating the biological, cognitive, and social dimensions of life into a science of sustainability. New York: Doubleday.

Eisenbarth, S. C., & Flavell, R. A. (2009).Innate instruction of adaptive immunity revisited: The inflammasome. EMBO Molecular Medicine, 1, 92-98.

Koyama, S., Ishii, K. J., Coban, C., & Akira, S. (2008). Innate immune response to viral infection. Cytokine, 43, 336-341.

Macal, C. M., & North, M. J. (2006). Tutorial on agent-based modeling and simulation, Part 2: How to model with agents. Proceedings of the 2006 Winter Simulation Conference (IEEE), Orlando, FL, 73-83.

McMichael, D., & Jarrad, G. (2005). Grammatical methods for situation and threat analysis. Proceedings of 8th International Conference on Information Fusion, Philadelphia, USA.

Medzhitov, R., & Janeway, Ch. (2000). Innate immunity. New England Journal of Medicine, 343, 338-344.

Miller, J. H., & Page, S. E. (2007). Complex adaptive systems: An introduction to computational models of social life. Princeton, NJ: Princeton University Press.

Moffat, J., Smith, M. J., & Witty, S. (2006). Emergent behavior: Theory and experimentation using the MANA model. Journal of Applied Mathematics and Decision Sciences, 1-13.

flotiuiiJKM. Vil Jtt

References