Download PDF by Roland Ewald: Automatic Algorithm Selection for Complex Simulation

By Roland Ewald

ISBN-10: 383481542X

ISBN-13: 9783834815422

To pick the main compatible simulation set of rules for a given activity is usually tricky. this can be because of difficult interactions among version positive factors, implementation info, and runtime atmosphere, that could strongly impact the final functionality. an automatic choice of simulation algorithms helps clients in developing simulation experiments with out tough professional wisdom on simulation.   Roland Ewald analyzes and discusses latest methods to resolve the set of rules choice challenge within the context of simulation. He introduces a framework for computerized simulation set of rules choice and describes its integration into the open-source modelling and simulation framework James II. Its choice mechanisms may be able to focus on 3 events: no previous wisdom is offered, the effect of challenge good points on simulator functionality is unknown, and a courting among challenge gains and set of rules functionality should be proven empirically. the writer concludes with an experimental evaluate of the built tools.

Show description

Read or Download Automatic Algorithm Selection for Complex Simulation Problems PDF

Best nonfiction_4 books

Additional resources for Automatic Algorithm Selection for Complex Simulation Problems

Example text

3, p. 30) is called a model. Following this nomenclature, the approximation form f (x) = α · x + β would be a model of all linear functions, with α and β being its parameters. To avoid confusion with simulation models, such models will explicitly called approximation models in the following. Each S ∈ S0 can be regarded as a hypothesis with respect to the suitability of the available algorithms for given features f ∈ F and user criteria w ∈ Rm . The restriction of the hypothesis space to S0 introduces the so-called approximation error, or bias, since the overall best selection mapping S∗ is not necessarily in S0 .

1 (Algorithm Selection Problem (ASP)) Let there be a problem space P, a feature space F, an algorithm space A, a criteria space Rn , a performance measure space Rn , a feature extraction mapping F : P → F, a performance mapping p : A × P → Rn , and a norm ||p(a, x)|| = g(p(a, x), w) with algorithm a ∈ A, problem x ∈ P, and user criterion w ∈ Rn . Determine a selection mapping S : F × Rn → A. 1 This reflects Rice’s basic definition (definition A, [272, p. 68]) but extends it by user criteria and feature extraction, which Rice only used in later definitions of ASP sub-problems.

Abstract machine models have been proposed that resemble modern computers in that they have RAM, are networked with other machines to exchange messages, and are interacting with a user [315]. Other approaches even extend the original definitions toward quantum computers [58] or biologically inspired computing [266]. Besides worst and average case analysis, Guo mentions the analysis of algorithm performance on certain sub-classes P0 ⊂ P [116, p. 20]. Rice hints at some problems regarding the characterization of selection mappings that could be investigated by complexity theory [272, p.

Download PDF sample

Automatic Algorithm Selection for Complex Simulation Problems by Roland Ewald

by William

Rated 4.07 of 5 – based on 5 votes