Proceedings of the AAAI Conference on Artificial Intelligence
Feature selection has been shown to be beneficial for many data mining and machine learning tasks... more Feature selection has been shown to be beneficial for many data mining and machine learning tasks, especially for big data analytics. Mutual Information (MI) is a well-known information-theoretic approach used to evaluate the relevance of feature subsets and class labels. However, estimating high-dimensional MI poses significant challenges. Consequently, a great deal of research has focused on using low-order MI approximations or computing a lower bound on MI called Variational Information (VI). These methods often require certain assumptions made on the probability distributions of features such that these distributions are realistic yet tractable to compute. In this paper, we reveal two sets of distribution assumptions underlying many MI and VI based methods: Feature Independence Distribution and Geometric Mean Distribution. We systematically analyze their strengths and weaknesses and propose a logical extension called Arithmetic Mean Distribution, which leads to an unbiased and n...
The version presented here may differ from the published version. If citing, you are advised to c... more The version presented here may differ from the published version. If citing, you are advised to consult the published version for pagination, volume/issue and date of publication 1 Norm talk and human cooperation: Can we talk ourselves into cooperation?
The sustainable use of common pool resources has become a significant global challenge. It is now... more The sustainable use of common pool resources has become a significant global challenge. It is now widely accepted that specific mechanisms such as community-based management strategies, institutional responses such as resource privatization, information availability and emergent social norms can be used to constrain individual 'harvesting' to socially optimal levels. However, there is a paucity of research focused specifically on aligning profitability and sustainability goals. In this paper, an integrated mathematical model of a common pool resource game is developed to explore the nexus between the underlying costs and benefits of harvesting decisions and the sustainable level of a shared, dynamic resource. We derive optimal harvesting efforts analytically and then use numerical simulations to show that individuals in a group can learn to make harvesting decisions that lead to the globally optimal levels. Individual agents make their decision based on signals received and a trade-off between economic and ecological sustainability. When the balance is weighted towards profitability, acceptable economic and social outcomes emerge. However, if individual agents are solely driven by profit, the shared resource is depleted in the long run-sustainability is possible despite some greed, but too much will lead to over-exploitation. The sustainable use of environmental, social and technical resources has become a significant global challenge 1, 2. Resource misuse, such as over-fishing 3-5 or deforestation 6-8 can potentially result in supply problems and lead to both economic and ecological damage. When the harvesting (or use of) a shared social-economic resource diminishes the value of the resource for other users (negative externality), and it is difficult to control access to the resource in the absence of well-defined property rights (non-excludability), the resource is typically referred to as a common pool resource (CPR) 9-12. CPR systems are characterized by a social dilemma-the tragedy of the commons 13-15. That is, the goal of an independently-acting individual is to maximize their use of the resource (gain higher portions of the harvest). However, if all individuals restrained their use of the resource, contrary to their selfish motivations, it should be possible to maintain the resource at a sustainable level, benefiting the population as a whole. An individual's selfish motivations to reap bigger profits manifest in the implicit assumption that investing more effort into harvesting will gain a larger proportion of the harvest and thus a higher profit, however, this proportional gains assumption is never expressed explicitly 16, 17. There is a large body of literature describing the management and governance of CPR systems. Perhaps most famous is the pioneering work of Elinor Ostrom 9, 15, 18 , who identified the benefits of managing the commons de-centrally and documented design principles for stable resource management. This work led to substantial related research in the field 19-22 , in laboratory settings 23-26 , as well as via simulation experiments 17, 27-30. Consequently, a number of external factors have been signalled as acting as drivers for cooperation in the commons, including: communication between individuals 21, 31-33 ; punishment of defectors 26, 34-37 ; reward 38-40 ; trust 14, 41, 42 ; social norms 22, 35, 43 ; and explicit consideration of the future 25, 44, 45 .
7th International Conference on Information and Automation for Sustainability, 2014
ABSTRACT Selecting the best algorithm for a given optimization problem is non-trivial due to larg... more ABSTRACT Selecting the best algorithm for a given optimization problem is non-trivial due to large number of existing algorithms and high complexity of problems. A possible way to tackle this challenge is to attempt to understand the problem complexity. Fitness Landscape Analysis (FLA) metrics are widely used techniques to extract characteristics from problems. Based on the extracted characteristics, machine learning methods are employed to select the optimal algorithm for a given problem. Therefore, the accuracy of the algorithm selection framework heavily relies on the choice of FLA metrics. Although researchers have paid great attention to designing FLA metrics to quantify the problem characteristics, there is still no agreement on which combination of FLA metrics should be employed. In this paper, we present some well-performed FLA metrics, discuss their contributions and limitations in detail, and map each FLA metric to the captured problem characteristics. Moreover, computational complexity of each FLA metric is carefully analysed. We propose two criteria to follow when selecting FLA metrics. We hope our work can help researchers identify the best combination of FLA metrics.
Proceedings of the 5th International Workshop on Middleware For Grid Computing Held at the Acm Ifip Usenix 8th International Middleware Conference, 2007
Most algorithms developed for scheduling applications on global Grids focus on a single Quality o... more Most algorithms developed for scheduling applications on global Grids focus on a single Quality of Service (QoS) parameter such as execution time, cost or total data transmission time. However, if we consider more than one QoS parameter (eg. execution cost and time may be in conflict) then the problem becomes more challenging. To handle such scenarios, it is convenient to use heuristics rather than a deterministic algorithm. In this paper we have proposed a workflow execution planning approach using Multiobjective Differential Evolution (MODE). Our goal was to generate a set of trade-off schedules according to two user specified QoS requirements (time and cost). The alternative tradeoff solutions offer more flexibility to users when estimating their QoS requirements of workflow executions. We have compared our results with two baseline multiobjective evolutionary algorithms. Simulation results show that our modified MODE is able to find a comparatively better spread of compromise solutions.
Proceedings of the 9th International Conference on Autonomous Agents and Multiagent Systems Volume 1 Volume 1, 2010
Understanding how cooperative behaviour emerges within a population of individuals has been the f... more Understanding how cooperative behaviour emerges within a population of individuals has been the focus of a great deal of research in the multi-agent systems community. In this paper, we examine the effectiveness of two different learning mechanisms-an evolutionary-based technique and a social imitation technique-in promoting and maintaining cooperation in the spatial N-player Iterated Prisoner's Dilemma (NIPD) game. Comprehensive Monte Carlo simulation experiments show that both mechanisms are able to evolve high levels of cooperation in the NIPD despite the diminished impact of direct reciprocation. However, the performance of evolutionary learning is significantly better than social learning, especially for larger population sizes. Our conclusion implies that when designing autonomous agents situated in complex environments, the use of evolutionarybased adaptation mechanisms will help realising efficient collective actions.
2008 Ieee Congress on Evolutionary Computation, 2008
Page 1. An analysis of the effects of clustering in graph-based evolutionary algorithms Cherhan F... more Page 1. An analysis of the effects of clustering in graph-based evolutionary algorithms Cherhan Foo and Michael Kirley Abstract Recently, there has been increased interest in com-bining work from the complex networks domain ...
Ieee Congress on Evolutionary Computation, Jul 18, 2010
The problem of evolving and maintaining cooperation in both ecological and artificial multi-agent... more The problem of evolving and maintaining cooperation in both ecological and artificial multi-agent systems has intrigued scientists for decades. In this paper, we present an evolutionary game model that combines direct and spatial reciprocity to investigate the effectiveness of two different learning mechanisms used to promote cooperative behaviour in a social dilemma game - the N-player Iterated Prisoner's Dilemma (NIPD).
Ieee Wic Acm International Conference on Intelligent Agent Technology, Oct 19, 2005
In this study, we investigate the underlying population dynamics when a group of agents compete f... more In this study, we investigate the underlying population dynamics when a group of agents compete for a finite, non-stationary resource. Based on a hybridized binary choice resource allocation game, at each time step individual agents make a decision whether to access the resource based on their own adaptive strategy and the aggregate history of previous resource utilization data. Agents, who are able to correctly identify the balance between supply and demand at the current time-step, are rewarded. However, agents who make an incorrect decision are penalized. Extensive numerical simulations show that the transient and long-run aggregate properties of the systems are dependent upon the rate of change of the resource availability as well as the heterogeneous decision making strategies adopted by the agent population.
We examine the (in)compatibility of diversity and sense of community by means of agent-based mode... more We examine the (in)compatibility of diversity and sense of community by means of agent-based models based on the well-known Schelling model of residential segregation and Axelrod model of cultural dissemination. We find that diversity and highly clustered social networks, on the assumptions of social tie formation based on spatial proximity and homophily, are incompatible when agent features are immutable, and this holds even for multiple independent features. We include both mutable and immutable features into a model that integrates Schelling and Axelrod models, and we find that even for multiple independent features, diversity and highly clustered social networks can be incompatible on the assumptions of social tie formation based on spatial proximity and homophily. However, this incompatibility breaks down when cultural diversity can be sufficiently large, at which point diversity and clustering need not be negatively correlated. This implies that segregation based on immutable characteristics such as race can possibly be overcome by sufficient similarity on mutable characteristics based on culture, which are subject to a process of social influence, provided a sufficiently large "scope of cultural possibilities" exists.
... of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard S... more ... of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen University of Dortmund, Germany Madhu Sudan Massachusetts ... by the four keynote speak-ers, and in addition, we were also fortunate to have Dipankar Dasgupta present a ...
XCS is an accuracy-based machine learning technique, which combines reinforcement learning and ev... more XCS is an accuracy-based machine learning technique, which combines reinforcement learning and evolutionary algorithms to evolve a set of classifiers (or rules) for pattern classification tasks. In this paper, we investigate the effects of alternative feature space partitioning techniques in a multiple population island-based parallel XCS. Here, each of the isolated populations evolve rules based on a subset of the features. The behavior of the multiple population model is carefully analyzed and compared with the original XCS using the Boolean logic multiplexer problem as a test case. Simulation results show that our multiple population XCS produced better performance and better generalization than the single population XCS model, especially when the problem increased in size. A caveat, however, is that the effectiveness of the model was dependent upon the feature space partitioning strategy used.
This paper describes an online collaborative learning environment - Speclad - for problem and pee... more This paper describes an online collaborative learning environment - Speclad - for problem and peer based learning. Speclad was originally developed as a platform to support student learning when constructing object-oriented class diagrams in software engineering and information system development. A number of enhancements have recently been introduced, which extends the functionality of the Speclad environment, including: the use of
Proceedings of the AAAI Conference on Artificial Intelligence
Feature selection has been shown to be beneficial for many data mining and machine learning tasks... more Feature selection has been shown to be beneficial for many data mining and machine learning tasks, especially for big data analytics. Mutual Information (MI) is a well-known information-theoretic approach used to evaluate the relevance of feature subsets and class labels. However, estimating high-dimensional MI poses significant challenges. Consequently, a great deal of research has focused on using low-order MI approximations or computing a lower bound on MI called Variational Information (VI). These methods often require certain assumptions made on the probability distributions of features such that these distributions are realistic yet tractable to compute. In this paper, we reveal two sets of distribution assumptions underlying many MI and VI based methods: Feature Independence Distribution and Geometric Mean Distribution. We systematically analyze their strengths and weaknesses and propose a logical extension called Arithmetic Mean Distribution, which leads to an unbiased and n...
The version presented here may differ from the published version. If citing, you are advised to c... more The version presented here may differ from the published version. If citing, you are advised to consult the published version for pagination, volume/issue and date of publication 1 Norm talk and human cooperation: Can we talk ourselves into cooperation?
The sustainable use of common pool resources has become a significant global challenge. It is now... more The sustainable use of common pool resources has become a significant global challenge. It is now widely accepted that specific mechanisms such as community-based management strategies, institutional responses such as resource privatization, information availability and emergent social norms can be used to constrain individual 'harvesting' to socially optimal levels. However, there is a paucity of research focused specifically on aligning profitability and sustainability goals. In this paper, an integrated mathematical model of a common pool resource game is developed to explore the nexus between the underlying costs and benefits of harvesting decisions and the sustainable level of a shared, dynamic resource. We derive optimal harvesting efforts analytically and then use numerical simulations to show that individuals in a group can learn to make harvesting decisions that lead to the globally optimal levels. Individual agents make their decision based on signals received and a trade-off between economic and ecological sustainability. When the balance is weighted towards profitability, acceptable economic and social outcomes emerge. However, if individual agents are solely driven by profit, the shared resource is depleted in the long run-sustainability is possible despite some greed, but too much will lead to over-exploitation. The sustainable use of environmental, social and technical resources has become a significant global challenge 1, 2. Resource misuse, such as over-fishing 3-5 or deforestation 6-8 can potentially result in supply problems and lead to both economic and ecological damage. When the harvesting (or use of) a shared social-economic resource diminishes the value of the resource for other users (negative externality), and it is difficult to control access to the resource in the absence of well-defined property rights (non-excludability), the resource is typically referred to as a common pool resource (CPR) 9-12. CPR systems are characterized by a social dilemma-the tragedy of the commons 13-15. That is, the goal of an independently-acting individual is to maximize their use of the resource (gain higher portions of the harvest). However, if all individuals restrained their use of the resource, contrary to their selfish motivations, it should be possible to maintain the resource at a sustainable level, benefiting the population as a whole. An individual's selfish motivations to reap bigger profits manifest in the implicit assumption that investing more effort into harvesting will gain a larger proportion of the harvest and thus a higher profit, however, this proportional gains assumption is never expressed explicitly 16, 17. There is a large body of literature describing the management and governance of CPR systems. Perhaps most famous is the pioneering work of Elinor Ostrom 9, 15, 18 , who identified the benefits of managing the commons de-centrally and documented design principles for stable resource management. This work led to substantial related research in the field 19-22 , in laboratory settings 23-26 , as well as via simulation experiments 17, 27-30. Consequently, a number of external factors have been signalled as acting as drivers for cooperation in the commons, including: communication between individuals 21, 31-33 ; punishment of defectors 26, 34-37 ; reward 38-40 ; trust 14, 41, 42 ; social norms 22, 35, 43 ; and explicit consideration of the future 25, 44, 45 .
7th International Conference on Information and Automation for Sustainability, 2014
ABSTRACT Selecting the best algorithm for a given optimization problem is non-trivial due to larg... more ABSTRACT Selecting the best algorithm for a given optimization problem is non-trivial due to large number of existing algorithms and high complexity of problems. A possible way to tackle this challenge is to attempt to understand the problem complexity. Fitness Landscape Analysis (FLA) metrics are widely used techniques to extract characteristics from problems. Based on the extracted characteristics, machine learning methods are employed to select the optimal algorithm for a given problem. Therefore, the accuracy of the algorithm selection framework heavily relies on the choice of FLA metrics. Although researchers have paid great attention to designing FLA metrics to quantify the problem characteristics, there is still no agreement on which combination of FLA metrics should be employed. In this paper, we present some well-performed FLA metrics, discuss their contributions and limitations in detail, and map each FLA metric to the captured problem characteristics. Moreover, computational complexity of each FLA metric is carefully analysed. We propose two criteria to follow when selecting FLA metrics. We hope our work can help researchers identify the best combination of FLA metrics.
Proceedings of the 5th International Workshop on Middleware For Grid Computing Held at the Acm Ifip Usenix 8th International Middleware Conference, 2007
Most algorithms developed for scheduling applications on global Grids focus on a single Quality o... more Most algorithms developed for scheduling applications on global Grids focus on a single Quality of Service (QoS) parameter such as execution time, cost or total data transmission time. However, if we consider more than one QoS parameter (eg. execution cost and time may be in conflict) then the problem becomes more challenging. To handle such scenarios, it is convenient to use heuristics rather than a deterministic algorithm. In this paper we have proposed a workflow execution planning approach using Multiobjective Differential Evolution (MODE). Our goal was to generate a set of trade-off schedules according to two user specified QoS requirements (time and cost). The alternative tradeoff solutions offer more flexibility to users when estimating their QoS requirements of workflow executions. We have compared our results with two baseline multiobjective evolutionary algorithms. Simulation results show that our modified MODE is able to find a comparatively better spread of compromise solutions.
Proceedings of the 9th International Conference on Autonomous Agents and Multiagent Systems Volume 1 Volume 1, 2010
Understanding how cooperative behaviour emerges within a population of individuals has been the f... more Understanding how cooperative behaviour emerges within a population of individuals has been the focus of a great deal of research in the multi-agent systems community. In this paper, we examine the effectiveness of two different learning mechanisms-an evolutionary-based technique and a social imitation technique-in promoting and maintaining cooperation in the spatial N-player Iterated Prisoner's Dilemma (NIPD) game. Comprehensive Monte Carlo simulation experiments show that both mechanisms are able to evolve high levels of cooperation in the NIPD despite the diminished impact of direct reciprocation. However, the performance of evolutionary learning is significantly better than social learning, especially for larger population sizes. Our conclusion implies that when designing autonomous agents situated in complex environments, the use of evolutionarybased adaptation mechanisms will help realising efficient collective actions.
2008 Ieee Congress on Evolutionary Computation, 2008
Page 1. An analysis of the effects of clustering in graph-based evolutionary algorithms Cherhan F... more Page 1. An analysis of the effects of clustering in graph-based evolutionary algorithms Cherhan Foo and Michael Kirley Abstract Recently, there has been increased interest in com-bining work from the complex networks domain ...
Ieee Congress on Evolutionary Computation, Jul 18, 2010
The problem of evolving and maintaining cooperation in both ecological and artificial multi-agent... more The problem of evolving and maintaining cooperation in both ecological and artificial multi-agent systems has intrigued scientists for decades. In this paper, we present an evolutionary game model that combines direct and spatial reciprocity to investigate the effectiveness of two different learning mechanisms used to promote cooperative behaviour in a social dilemma game - the N-player Iterated Prisoner's Dilemma (NIPD).
Ieee Wic Acm International Conference on Intelligent Agent Technology, Oct 19, 2005
In this study, we investigate the underlying population dynamics when a group of agents compete f... more In this study, we investigate the underlying population dynamics when a group of agents compete for a finite, non-stationary resource. Based on a hybridized binary choice resource allocation game, at each time step individual agents make a decision whether to access the resource based on their own adaptive strategy and the aggregate history of previous resource utilization data. Agents, who are able to correctly identify the balance between supply and demand at the current time-step, are rewarded. However, agents who make an incorrect decision are penalized. Extensive numerical simulations show that the transient and long-run aggregate properties of the systems are dependent upon the rate of change of the resource availability as well as the heterogeneous decision making strategies adopted by the agent population.
We examine the (in)compatibility of diversity and sense of community by means of agent-based mode... more We examine the (in)compatibility of diversity and sense of community by means of agent-based models based on the well-known Schelling model of residential segregation and Axelrod model of cultural dissemination. We find that diversity and highly clustered social networks, on the assumptions of social tie formation based on spatial proximity and homophily, are incompatible when agent features are immutable, and this holds even for multiple independent features. We include both mutable and immutable features into a model that integrates Schelling and Axelrod models, and we find that even for multiple independent features, diversity and highly clustered social networks can be incompatible on the assumptions of social tie formation based on spatial proximity and homophily. However, this incompatibility breaks down when cultural diversity can be sufficiently large, at which point diversity and clustering need not be negatively correlated. This implies that segregation based on immutable characteristics such as race can possibly be overcome by sufficient similarity on mutable characteristics based on culture, which are subject to a process of social influence, provided a sufficiently large "scope of cultural possibilities" exists.
... of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard S... more ... of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen University of Dortmund, Germany Madhu Sudan Massachusetts ... by the four keynote speak-ers, and in addition, we were also fortunate to have Dipankar Dasgupta present a ...
XCS is an accuracy-based machine learning technique, which combines reinforcement learning and ev... more XCS is an accuracy-based machine learning technique, which combines reinforcement learning and evolutionary algorithms to evolve a set of classifiers (or rules) for pattern classification tasks. In this paper, we investigate the effects of alternative feature space partitioning techniques in a multiple population island-based parallel XCS. Here, each of the isolated populations evolve rules based on a subset of the features. The behavior of the multiple population model is carefully analyzed and compared with the original XCS using the Boolean logic multiplexer problem as a test case. Simulation results show that our multiple population XCS produced better performance and better generalization than the single population XCS model, especially when the problem increased in size. A caveat, however, is that the effectiveness of the model was dependent upon the feature space partitioning strategy used.
This paper describes an online collaborative learning environment - Speclad - for problem and pee... more This paper describes an online collaborative learning environment - Speclad - for problem and peer based learning. Speclad was originally developed as a platform to support student learning when constructing object-oriented class diagrams in software engineering and information system development. A number of enhancements have recently been introduced, which extends the functionality of the Speclad environment, including: the use of
Uploads
Papers by Michael Kirley