Modeling Natural Immune Systems using Complex Systems Theories
Team Members: Asst Prof. Tay Joc Cing (PI), Hidefumi Sawai (co-PI)
Collaborators: Kansai Advanced Research Center
Police and Thief is a children’s game based on symmetrical pursuit and evasion. In the projects described here, control programs for mobile agents (simulated vehicles) are evolved based on their skill at the game of Police and Thief. A player’s fitness is determined by how well it performs when placed in competition with several opponents chosen randomly from the coevolving population of players. In the beginning, the quality of play is very poor. Then slightly better strategies begin to exploit the weaknesses of others. Through evolution, guided by competitive fitness, increasingly better strategies emerge over time.
Coevolutionary Algorithms (CEAs) are believed to be useful in tackling problems involving infinite search spaces as well as in problems for which no intrinsic objective measures exists. However, the dynamics of CEAs have yet to be fully understood, as the algorithms frequently exhibit counter intuitive behaviors.
The game of coevolutionary pursuit and evasion is categorized under competitive interaction in CEAs. In this research, we have successfully implemented this application using CGP. By playing the game in an arena involving obstacles and boundaries, we make observations regarding the mobile agents’ behavior in a coevolutionary environment. This is a step towards a better understanding of how CEAs can be used to model cooperative and competitive behaviours.
For more information, please refer to the poster and the website.
Grid-enabled Optimization Environment for Complex Engineering Problems
Team Members: Asst Prof Ong Yew-Soon (PI), Lim Dudy (co-PI), Ng Hee Khiang Lee Bu-Sung (co-PI)
Collaborators: Honda Research Institute Europe GmbH
The primary focus of this research proposal is to develop a Grid enabled evolutionary algorithm for solving computationally expensive problem in complex engineering design optimization. These involve the design and development of novel tools for seamless ‘Gridification’ of existing applications, i.e., EA algorithms and analysis/simulations codes, as Grid resources as well as mechanisms for consumption of Grid resources.
Models of Cooperation and Conflict for Immune System Response
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Bui Van Quang (Student)
Collaborators: Asst Prof. Raymond Wong (City University of Hong Kong)
The current simulation approach can be further extended by considering the interaction between the HIV virus and the immune system as a co-evolutionary process within a game theoretic framework. In other words, the virus and various component cells of the immune system; including the antibodies and T-helper cells, form different sub-populations which interact and compete with each other to maximize an associated payoff function. Co-evolutionary process, which describes the evolution and dominance of different agent types in their respective sub-populations, has been studied from the perspective of economics and mathematical biology, and forms a particularly suitable model for our current task. This is in view of the existence, also in our case, of different agent types in each sub-population of our model (e.g., the various mutated forms of the HIV virus, and the corresponding diverse types of antibodies in response to the high mutation rate of the virus), which fits neatly into the co-evolutionary framework. In this research we wish to investigate the use of cooperative or uncooperative combinatorial game theories to model the dynamics of immune system responses to viral infection with an objective towards deriving emergent and learnable behaviours.
Modeling Natural Immune Systems using Complex Systems Theories
Team Members: Asst Prof. Tay Joc Cing (PI), Ms. Guo Zaiyi (Student),
Collaborators: Assoc Prof Kwoh Chee Keong (Deputy Director, BMERC, School of Computer Engineering), Dr. Nick Paton (Department of Infectious Disease, Tan Tock Seng Hospital)
The goal of natural immune systems (or NIS) modeling is to provide powerful mathematical and computational tools to facilitate immunology study. This is in contrast to artificial immune systems (or AIS), which is defined as “intelligent methodologies inspired by the immune system toward real-world problem solving”. The typical expected outcomes of NIS modeling are:
- Virtual
Laboratories. The models serve as virtual laboratories in which
numerical experiments can be performed. In real experiments, some analyses
are not feasible, such as due to the long time span, or not even possible,
such as altering certain parameters to gain insights of their impact and
relative importance.
- Hypothesis
Verification. A hypothesis is a tentative explanation for an observation,
a phenomenon or a scientific problem that will be tested by further
investigation. The hypothesis in immunology can be an intuitive guess of
the mechanisms or the cause-effect relationship based on observations.
Hypothesis can be experimentally ‘tested’ by comparing its predictions
with reality. Since the biological
systems are complex and difficult to control in nature, the successful
comparison results can give only statistical support to the hypothesis.
The modeling approach can help one step further. By establishing a model,
one can ‘verify’ whether the assumptions and explanations really work in a
manner expected by intuition and lead to the observed or predicted
phenomenon.
Immune Response Model Verification using CAFISS
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Han Hann Kwang (Student)
Collaborators: Dr. Nick Paton (Department of Infectious Disease,Tan Tock Seng Hospital)
The immune system has been difficult to model due to the complexities arising from the interactions of the cells. Even though there are many different methods to simulate the immune system, they are found to be inadequate and limit the behavior response of the immune system entities. Complex Adaptive System or CAS provides a novel way of modeling the entities with emergent behaviors. The immune system is made up of relatively simple agents, operating with highly complex and coordinated behaviors. This makes the immune system particularly suited to being modeled using CAS. This paper will discuss how the immune system is model and particularly against HIV attack. Three theories of HIV infection are implemented using CAFISS, a CAS-based multiagent simulation platform for immune systems.
Developing Visualization Algorithms for Evolutionary Computation
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Ong Zi Xuan (Student)
Evolutionary algorithms work in an algorithmically simple manner but produce a vast amount of data. The extraction of useful information to gain further insight into state and course of the algorithm is a non-trivial task. In this project we wish to develop a set of standard visualization techniques for different data types and time frames of the evolutionary algorithm. The methods are selected according to their usefulness for real world applications, and tested during the solution of some complex real world optimization problems.
A Genetic Programming Platform for Visualizing Two-Player Games
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Chan Chee Siong (Student)
Collaborators: Prof. Hidefumi Sawai (Graduate School of Kobe University)
Police and Thief is a children’s game based on symmetrical pursuit and evasion. In the projects described here, control programs for mobile agents (simulated vehicles) are evolved based on their skill at the game of Police and Thief. A player’s fitness is determined by how well it performs when placed in competition with several opponents chosen randomly from the coevolving population of players. In the beginning, the quality of play is very poor. Then slightly better strategies begin to exploit the weaknesses of others. Through evolution, guided by competitive fitness, increasingly better strategies emerge over time.
Implementation of a Viral-Evolutionary Genetic Algorithm to solve the Open-Shop Scheduling Problem
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Kuah Shylin (Student)
The open-shop scheduling problem (or OSSP) is an important practical scheduling problem. It arises in an environment where there is a collection of jobs to perform on a set of machines. In the j x m OSSP, there are j jobs and m machines. Each job consists of a set of operations n that are processed on a predefined number of machines for different amount of times. There are no restrictions with regard to the order of operations of each job. For instances, in automotive repair shop, a typical job might involve the operation ‘fix-brakes’, ‘change-tyres’ and ‘spray-paint’. The operations can be performed in any order. However, this problem is typical NP-hard, it is impossible to find optimum solutions without using appropriate local search approaches. For example, on the 5 jobs x 5 machines OSSP, there are 25! different jobs orderings which is approximately 1.5x1025 different schedules.
Development of a Hybrid Evolutionary Optimization Algorithm for the Flexible Job-Shop Problem
Team Members: Asst Prof. Tay Joc Cing (PI), Mr. Tan Boon Tai (Student)
Classical methods for optimization suffer from cumbersome models and non-polynomial worst-case complexities. In addition, there are difficulties in meeting the assumptions of differentiability and convexity in the constraints specified. In this project, we wish to
- Understand evolutionary computation and linear
programming.
- Develop a hybrid algorithm in which an evolutionary
algorithm is used to remove local optima, and in the second stage, a
classical optimization technique such as simplex is used to improve the
local convergence.
- Survey other works involving hybrid optimization
algorithms.
- Do performance comparison studies and evaluate
results.
Approximating the travelling salesman problem with an extended Artificial Immune Systems Paradigm
Team Members: Asst Prof. Tay Joc Cing (PI), Ms. Serene Wong (Student)
Artificial Immune Systems offers a new heuristic to solve optimization problems through the evolution of antibodies. In this project, we first apply this technique to approximate the geometric TSP. Afterwhich, we extend the technique to include interactions with T cells and Antigen presenting cells. Performance of the new algorithm is then compared against the basic version.
Simulating the Evolution of Mating Preferences through CAFISS
Team Members: Asst Prof. Tay Joc Cing (PI),Mr. Chan Yuh Miin (Student)
Complex Adaptive Systems (or CAS) [Holland], is a behavioural evolution model for systems with emergent behaviours due to complex aggregation and collective interactions. It is used as a framework to simulate natural phenomena so as to understand the underlying governing mechanisms. This model is particularly suited for implementing agents that are capable of adapting under complex environments. In this project, we seek to develop a framework for adaptive agents. The parts of this agent will comprise of (1) a rule-based agent model, (2) a Credit-assignment mechanism and (3) an evolution algorithm for generating new planning rules. The evolutionary algorithm used in (3) will be based on an available GA library.
Evaluating the Baldwin Effect of Learning on Evolution
Team Members: Asst Prof. Tay Joc Cing
Verification of Topological and Activation Schema Effects on HIV Simulation
Team Members: Asst Prof. Tay Joc Cing
Hypothesis Verification of Dengue Pathogenesis using CAFISS
Team Members: Asst Prof. Tay Joc Cing
|
