DG2: A Faster and More Accurate Differential Grouping for Large-Scale Black-Box Optimization

Abstract

Link to publication on Research at Birmingham portal General rights Unless a licence is specified above, all rights (including copyright and moral rights) in this document are retained by the authors and/or the copyright holders. The express permission of the copyright holder must be obtained for any use of this material other than for purposes permitted by law. • Users may freely distribute the URL that is used to identify this publication. • Users may download and/or print one copy of the publication from the University of Birmingham research portal for the purpose of private study or non-commercial research. • User may use extracts from the document in line with the concept of 'fair dealing' under the Copyright, Designs and Patents Act 1988 (?) • Users may not further distribute the material nor use it for the purposes of commercial gain. Where a licence is displayed above, please note the terms and conditions of the licence govern your use of this document. When citing, please reference the published version. Take down policy While the University of Birmingham exercises care and attention in making items available there are rare occasions when an item has been uploaded in error or has been deemed to be commercially or otherwise sensitive.

A global information based adaptive threshold for grouping large scale optimization problems

“…The threshold parameter has a great influence on the decomposition accuracy of DG-based algorithms, there are three methods to set it up to now, including fixed threshold (FT) [9], function space based threshold (FST) [11] and computational roundoff error based threshold (CRET) [13]. In this part, we will review them in sequence.…”

“…Besides, its decomposition accuracy is also limited by the threshold setting method. After DG was proposed, many algorithms were developed to improved it, including global DG [11], extend DG [10], FII [12], DG2 [13] and VGDA [14]. The first two drawbacks have been remedied very well, but for threshold setting, some work still need to be done to improve it.…”

A Parallel Divide-and-Conquer-Based Evolutionary Algorithm for Large-Scale Optimization

“…The CEC'2010 large-scale optimization benchmark covers different degrees of separability and multi-modality, which are the main difficulties for large-scale optimization problems [59], and thus has been commonly used to verify the performance of large-scale optimization algorithms [17], [33], [60]. Given this, it is also adopted as the test suite in the empirical studies.…”

“…The on-line random grouping happens every iteration as [31] proved that randomly grouping more frequently could minimize the interdependencies among sub-problems. DC-DG decomposes the problem using the improved differential grouping strategy, i.e., DG2 [60], which actively analyzes the interdependencies among decision variables so that the decomposition can be made much more accurate. The only difference between the above three compared algorithms and their corresponding parallelized counterparts is that the former builds the objective function for sub-problems using Eq.…”

“…That is, the higher the interdependencies exist between two sub-problems, their objective functions are more likely to be inconsistent with f . Nevertheless, even the DG2 decomposition strategy has empirically shown the perfect decomposition, i.e., no interdependencies among sub-problems, on most problems in CEC'2010 test suite [60], the divergences between DC-DG and DC-DG-P on some problems are still significant. This phenomenon again stresses that the existing DC-based EAs are difficult to be parallelized, unless some compromises on the final solution qualities are made.…”