Matching inductive search bias and problem structure in continuous Estimation-of-Distribution Algorithms

Bosman, Peter A. N.; Grahl, John

doi:10.1016/j.ejor.2006.06.051

Cited by 44 publications

(31 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[10][11][12][13] The particular EA that we employ is iMAMaLGaM (incremental Multi-objective Adapted Maximum Likelihood Gaussian Model mixture) in which the underlying probabilistic model is a Gaussian mixture distribution. 4 In related work, iMAMaLGaM was shown to have excellent performance, converging to high-quality approximations of the optimal Pareto front on well-known benchmark problems.…”

Section: Multi-objective Optimization Algorithmmentioning

confidence: 99%

Smart grid initialization reduces the computational complexity of multi-objective image registration based on a dual-dynamic transformation model to account for large anatomical differences

Bosman

Alderliesten

2016

SPIE Proceedings

View full text Add to dashboard Cite

We recently demonstrated the strong potential of using dual-dynamic transformation models when tackling deformable image registration problems involving large anatomical differences. Dual-dynamic transformation models employ two moving grids instead of the common single moving grid for the target image (and single fixed grid for the source image). We previously employed powerful optimization algorithms to make use of the additional flexibility offered by a dualdynamic transformation model with good results, directly obtaining insight into the trade-off between important registration objectives as a result of taking a multi-objective approach to optimization. However, optimization has so far been initialized using two regular grids, which still leaves a great potential of dual-dynamic transformation models untapped: a-priori grid alignment with image structures/areas that are expected to deform more. This allows (far) less grid points to be used, compared to using a sufficiently refined regular grid, leading to (far) more efficient optimization, or, equivalently, more accurate results using the same number of grid points. We study the implications of exploiting this potential by experimenting with two new smart grid initialization procedures: one manual expert-based and one automated image-feature-based. We consider a CT test case with large differences in bladder volume with and without a multi-resolution scheme and find a substantial benefit of using smart grid initialization.

show abstract

Section: Multi-objective Optimization Algorithmmentioning

confidence: 99%

Smart grid initialization reduces the computational complexity of multi-objective image registration based on a dual-dynamic transformation model to account for large anatomical differences

Bosman

Alderliesten

2016

SPIE Proceedings

View full text Add to dashboard Cite

show abstract

“…the Gaussian distribution should fit the local neighborhood much better than in case of EMNA or CMA-ES, 2. it works with individuals marked only with select/discard labels, it does not need the fitness values for each of them (as is the case in [1]), 3. it estimates a 'reasonable' Gaussian even from a small number of individuals (much less than D(D + 3)/2 + 1 that are needed by [1]).…”

Section: Summary and Future Workmentioning

confidence: 99%

“…1(a) for an example of Estimation of Multivariate Normal distribution Algorithm (EMNA)). Without imposing limits on the minimal 'size' of the Gaussian, the variance of the distribution in the direction of the fitness function gradient quickly decreases, and the algorithm thus can get stuck even on the slope of the fitness function [3].…”

Section: Introductionmentioning

confidence: 99%

Estimation of fitness landscape contours in EAs

Pošík

2007

Proceedings of the 9th Annual Conference on Genetic and Evolutionary Computation

View full text Add to dashboard Cite

Evolutionary algorithms applied in real domain should profit from information about the local fitness function curvature. This paper presents an initial study of an evolutionary strategy with a novel approach for learning the covariance matrix of a Gaussian distribution. The learning method is based on estimation of the fitness landscape contour line between the selected and discarded individuals. The distribution learned this way is then used to generate new population members. The algorithm presented here is the first attempt to construct the Gaussian distribution this way and should be considered only a proof of concept; nevertheless, the empirical comparison on low-dimensional quadratic functions shows that our approach is viable and with respect to the number of evaluations needed to find a solution of certain quality, it is comparable to the state-of-the-art CMA-ES in case of sphere function and outperforms the CMA-ES in case of elliptical function.

show abstract

“…In [21] and [22], Adaptive Variance Scaling (AVS) and Correlation Triggered AVS (CT-AVS) were proposed to use along with normal pdf in IDEA. The essential idea of AVS is to scale covariance matrix with an adaptive (also positive) coefficient c AVS to help increase the area of exploration.…”

Section: ) Negative Variance Of Egnamentioning

confidence: 99%

Covariance matrix repairing in Gaussian based EDAs

Dong

Yao²

2007

2007 IEEE Congress on Evolutionary Computation

View full text Add to dashboard Cite

Abstract-Gaussian models are widely adopted in continuous Estimation of Distribution Algorithms (EDAs). In this paper, we analyze continuous EDAs and show that they don't always work because of computation error: covariance matrix of Gaussian model can be ill-posed and Gaussian based EDAs using full covariance matrix will fail under specific conditions. It is a universal problem that all existing Gaussian based EDAs using full covariance matrix suffer from. Through theoretical analysis with examples of simulated data and experiments, we show that the ill-posed covariance matrix strongly affects those EDAs. This paper proposes a Covariance Matrix Repairing (CMR) method to fix ill-posed covariance matrix. CMR significantly improves the robustness of EDAs. Even some EDA's performance that was previously thought inefficient can be improved surprisingly with the help of CMR. CMR can also guarantee those EDAs to be used with small scale of population (but still should be large enough to find the global optimum) to accelerate the convergence rate while maintaining the quality of solutions.

show abstract

Matching inductive search bias and problem structure in continuous Estimation-of-Distribution Algorithms

Cited by 44 publications

References 28 publications

Smart grid initialization reduces the computational complexity of multi-objective image registration based on a dual-dynamic transformation model to account for large anatomical differences

Smart grid initialization reduces the computational complexity of multi-objective image registration based on a dual-dynamic transformation model to account for large anatomical differences

Estimation of fitness landscape contours in EAs

Covariance matrix repairing in Gaussian based EDAs

Contact Info

Product

Resources

About