Overcoming hierarchical difficulty by hill-climbing the building block structure

Iclănzan, David; Dumitrescu, D.

doi:10.1145/1276958.1277199

Cited by 19 publications

(31 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[12]) but in optimisation the ability to produce new combinations of successful features is highly-desirable [21,77,80,81]. Mills [46,47,44] shows that the automatic discovery and utilisation of modular structure in an optimisation problem, as facilitated by learned associations, can be used to provide significant optimisation performance (see also [26,27,84,93]). This result thereby shows that a distributed optimisation process, based on nothing more than repeated relaxation of state configurations plus local selfish reinforcement of connections has the effect not only of creating an associative memory of its past local optimisation behaviour but also generalising its past behaviour and enabling superior optimisation and, in the context of a multi-agent system, global adaptation.…”

Section: Memory Optimisation and Generalisationmentioning

confidence: 99%

Global Adaptation in Networks of Selfish Components: Emergent Associative Memory at the System Scale

2011

View full text Add to dashboard Cite

In some circumstances complex adaptive systems composed of numerous self-interested agents can self-organise into structures that enhance global adaptation, efficiency or function. However, the general conditions for such an outcome are poorly understood and present a fundamental open question for domains as varied as ecology, sociology, economics, organismic biology and technological infrastructure design. In contrast, sufficient conditions for artificial neural networks to form structures that perform collective computational processes such as associative memory/recall, classification, generalisation and optimisation, are well-understood. Such global functions within a single agent or organism are not wholly surprising since the mechanisms (e.g. Hebbian learning) that create these neural organisations may be selected for this purpose, but agents in a multi-agent system have no obvious reason to adhere to such a structuring protocol or produce such global behaviours when acting from individual self-interest. However, Hebbian learning is actually a very simple and fully-distributed habituation or positive feedback principle. Here we show that when self-interested agents can modify how they are affected by other agents (e.g. when they can influence which other agents they interact with) then, in adapting these inter-agent relationships to maximise their own utility, they will necessarily alter them in a manner homologous with Hebbian learning. Multi-agent systems with adaptable relationships will thereby exhibit the same system-level behaviours as neural networks under Hebbian learning. For example, improved global efficiency in multi-agent systems can be explained by the inherent ability of associative memory to generalise by idealising stored patterns and/or creating new combinations of sub-patterns. Thus distributed multi-agent systems can spontaneously exhibit adaptive global behaviours in the same sense, and by the same mechanism, as the organisational principles familiar in connectionist models of organismic learning.Keywords: self-organisation, adaptive networks, Hebbian learning, multi-agent systems, social networks, emergent computation, games on networks. Selfish changes to connections and global adaptationOne of the key open questions in the field of artificial life and adaptive systems is how, if at all, it is possible that a complex system that is not evolved can exhibit adaptation or increased functionality without design. has the potential to play a role in the formation of pre-biotic organisations, or in moving from one level of biological organisation to another [42], for example, but theory to understand exactly what this means or how it might work is limited. In contrast, theory to understand distributed adaptive processes in neural networks is well-developed, but generally assumed to be relevant only to brains and nervous systems 'programmed' to exhibit such adaptation. Here we show that organisational principles familiar in learning neural networks emerge spontaneously in distributed...

show abstract

Section: Memory Optimisation and Generalisationmentioning

confidence: 99%

Global Adaptation in Networks of Selfish Components: Emergent Associative Memory at the System Scale

2011

View full text Add to dashboard Cite

show abstract

“…In addition, by altering the given examples of recursive instantiation to make a stochastic choice of lower-level (hyper-)heuristics, Hyperion can also be considered as a generation mechanism for strongly-typed genetic programming [32] in the domain of hyper-heuristics. Future work includes an investigation of the eect of recursion depth in the context of building-blocks in`hierarchical i ' functions [33]. There are also a number of aspects of the current framework implementation that we believe could be improved upon.…”

Section: Discussionmentioning

confidence: 99%

Hyperion – A Recursive Hyper-Heuristic Framework

Swan

Özcan

Kendall

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Hyper-heuristics are methodologies used to search the space of heuristics for solving computationally dicult problems. We describe an object-oriented domain analysis for hyper-heuristics that orthogonally decomposes the domain into generative policy components. The framework facilitates the recursive instantiation of hyper-heuristics over hyper-heuristics, allowing further exploration of the possibilities implied by the hyper-heuristic concept. We describe Hyperion, a Java TM class library implementation of this domain analysis.

show abstract

“…Elsewhere we have investigated landscapes in which associations with intermediate strengths find high-utility configurations that cannot be found with 'all-or-nothing' associations [19,20]. It is worth noting that the algorithm proposed in [16] is able to efficiently solve hierarchical problems with some similar abstractions, in particular by model building from information at local optima.…”

Section: Discussionmentioning

confidence: 99%

“…As noted, we use different temporal scales for ecosystem dynamics and association formation. In evolutionary computation, [16] uses results of multiple hill climbing runs to build a model of dependencies, which provides a similar timescale separation to successfully solve hierarchical problems. Memetic algorithms [17] also use search at two levels, but importantly, neither search process modifies the variational units for the other.…”

Section: Introductionmentioning

confidence: 99%