Automated Extraction of Problem Structure

Bucci, Anthony; Pollack, Jordan; Jong, Edwin de

doi:10.1007/978-3-540-24854-5_53

Cited by 37 publications

(25 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In some cases the objective error of a model approaches infinity, due to some inaccuracy of the model that causes positive or negative feedback. In these cases the model is discarded and the mean objective is calculated using only the remaining well behaved models 5 . Figures 3b and 3c report the mean number of target trials and model evaluations that have been conducted so far, at the end of that particular pass through the estimation phase, respectively.…”

Section: Parametric Identification Resultsmentioning

confidence: 99%

See 1 more Smart Citation

'Managed challenge' alleviates disengagement in co-evolutionary system identification

Bongard

Lipson

2005

Proceedings of the 7th Annual Conference on Genetic and Evolutionary Computation

View full text Add to dashboard Cite

In previous papers we have described a co-evolutionary algorithm (EEA), the estimation-exploration algorithm, that infers the hidden inner structure of systems using minimal testing. In this paper we introduce the concept of 'managed challenge' to alleviate the problem of disengagement in this and other co-evol-utionary algorithms. A known problem in co-evolutionary dynamics occurs when one population systematically outperforms the other, resulting in a loss of selection pressure for both populations. In system identification (which deals with determining the inner structure of a system using only input/output data), multiple trials (a test that causes the system to produce some output) on the system to be identified must be performed. When such trials are costly, this disengagement results in wasted data that is not utilized by the evolutionary process. Here we propose that data from futile interactions should be stored during disengagement and automatically re-introduced later, when the population re-engages: we refer to this as the test bank. We demonstrate that the advantage of the test bank is twofold: it allows for the discovery of more accurate models, and it reduces the amount of required training data for both parametric identification -parameterizing inner structureand symbolic identification -approximating inner structure using symbolic equations -of nonlinear systems.

show abstract

Section: Parametric Identification Resultsmentioning

confidence: 99%

“…5 A model may be well behaved but still inaccurate 1) Characterization of the target system. The target system is the same as that used in the previous section, the two-eyed monster.…”

Section: Symbolic Identificationmentioning

confidence: 99%

'Managed challenge' alleviates disengagement in co-evolutionary system identification

Bongard

Lipson

2005

Proceedings of the 7th Annual Conference on Genetic and Evolutionary Computation

View full text Add to dashboard Cite

show abstract

“…Using order theory, Bucci [6,7,8] has deduced that every interactive domain with binary outcomes possesses at least one coordinate system which spans the original domain's Pareto dominance ordering. Each coordinate axis corresponds to a dimension.…”

Section: Preliminariesmentioning

confidence: 99%

“…Recent work [6,7,8] has indicated that within interactive domains, there is an implicit set of yardsticks which might be called informative dimensions. Furthermore, a characterizing property of interactive domains is that they have more than one dimension.…”

Section: Introductionmentioning

confidence: 99%

A Simple Coevolution Archive Based on Bidirectional Dimension Extraction

Yang

Huang

Yang

2009

2009 International Conference on Artificial Intelligence and Computational Intelligence

View full text Add to dashboard Cite

A common technique in coevolution aimed at improving reliability is the use of an archive. Recent research has shown that within coevolutionary problem domains, there is an implicit set of informative dimensions which structures the evaluation space for individuals. According to above research, in this paper we propose a novel archive algorithm, which adopts a simple and feasible dimension extracting method, extracts dimension systems from both candidates and tests sides respectively, and select only high-performance individuals representing informative dimensions to reside in archives. Therefore, the proposed algorithm can prevent archives from growing overly larger while maintaining progress. Furthermore, as the conditions of dimension extraction are reduced, it is more widely applicable than other similar methods. Experimental results on test problems demonstrate the feasibility and validity of the proposed algorithm.

show abstract

“…Furthermore, co-evolutionary algorithms have also been used in structural and architectural design [13,14]. Contrary to conventional evolutionary systems, in which individuals are evaluated using a static quality or fitness metric, co-evolutionary systems consist of one or more populations in which individuals may influence the relative ranking of each other [15]. A new methodology based on co-evolution is presented herein for damage identification using minimal physical testing [11].…”

Section: Introductionmentioning

confidence: 99%

Co‐evolutionary algorithm for structural damage identification using minimal physical testing

Kouchmeshky

Aquino

Bongard

et al. 2006

Numerical Meth Engineering

View full text Add to dashboard Cite

SUMMARYThe problem of damage identification using minimum test data is studied in this work. Data sparsity in damage identification applications commonly results in inverse problems that are mathematically ill-posed (e.g. non-unique solutions). Although solution non-uniqueness may be addressed by performing multiple tests on a structure, it is not trivial to decide which tests to carry out given that actual physical testing is costly. This problem is addressed in this work through a new co-evolutionary algorithm that interactively searches for damage scenarios and optimum physical tests. The algorithm is composed of two stages: the estimation phase, which searches for damage scenarios that can predict current physical tests, and the exploration phase, which searches for tests that increase the level of information about the damaged system. The feasibility of the methodology is demonstrated using numerical examples.

show abstract

Automated Extraction of Problem Structure

Cited by 37 publications

References 6 publications

'Managed challenge' alleviates disengagement in co-evolutionary system identification

'Managed challenge' alleviates disengagement in co-evolutionary system identification

A Simple Coevolution Archive Based on Bidirectional Dimension Extraction

Co‐evolutionary algorithm for structural damage identification using minimal physical testing

Contact Info

Product

Resources

About