Parallel Black-Box Complexity With Tail Bounds

@article{Lehre2020ParallelBC,
  title={Parallel Black-Box Complexity With Tail Bounds},
  author={P. Lehre and Dirk Sudholt},
  journal={IEEE Transactions on Evolutionary Computation},
  year={2020},
  volume={24},
  pages={1010-1024}
}
We propose a new black-box complexity model for search algorithms evaluating <inline-formula> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> search points in parallel. The parallel unary unbiased black-box complexity gives lower bounds on the number of function evaluations <italic>every</italic> parallel unary unbiased black-box algorithm needs to optimize a given problem. It captures the inertia caused by offspring populations in evolutionary algorithms and the total… Expand
7 Citations
Blending Dynamic Programming with Monte Carlo Simulation for Bounding the Running Time of Evolutionary Algorithms
TLDR
The hybrid Monte Carlo dynamic programming approach is applied to a concatenated jump function and it is demonstrated how the obtained bounds can be used to gain a deeper understanding into parameter control schemes. Expand
Theory of Iterative Optimization Heuristics: From Black-Box Complexity over Algorithm Design to Parameter Control
TLDR
Improved bounds for the black-box complexity of the two best known benchmark problems in the theory of IOHs, OneMax and LeadingOnes are derived, and it is demonstrated how insights obtained from such black- box complexity studies can inspire the design of efficient optimization techniques. Expand
On the impact of the cutoff time on the performance of algorithm configurators
TLDR
This work evaluates the performance of a simple random local search configurator (ParamRLS) for tuning the neighbourhood size k of the RLSk algorithm and rigorously proves that ParamRLS-F efficiently tunes R LSk for Ridge* for any κ while ParamR LS-T requires at least quadratic κ. Expand
Illustrating the trade-off between time, quality, and success probability in heuristic search: a discussion paper
TLDR
It is suggested to extend the classically regarded fixed-target and fixed-budget analyses by a fixed-probability measure, and provide a first implementation towards a Fixed-Probability module within the IOHprofiler environment. Expand
Choosing the Right Algorithm With Hints From Complexity Theory
Choosing a suitable algorithm from the myriads of different search heuristics is difficult when faced with a novel optimization problem. In this work, we argue that the purely academic question ofExpand
Do additional optima speed up evolutionary algorithms?
Most runtime analyses of randomised search heuristics focus on the expected number of function evaluations to find a unique global optimum. We ask a fundamental question: if additional search pointsExpand
Optimal static mutation strength distributions for the (1 + λ) evolutionary algorithm on OneMax
TLDR
An algorithm is presented that computes the optimal static distributions, which describe the mutation operator used in the well-known simple (1+λ) evolutionary algorithm on a classic benchmark problem OneMax, and it is shown that, for large enough population sizes, such optimal distributions may be surprisingly complicated and counter-intuitive. Expand

References

SHOWING 1-10 OF 74 REFERENCES
Unbiased Black-Box Complexity of Parallel Search
TLDR
A new black-box complexity model for search algorithms evaluating λ search points in parallel that captures the inertia caused by offspring populations in evolutionary algorithms and the total computational effort in parallel metaheuristics is proposed. Expand
From black-box complexity to designing new genetic algorithms
TLDR
This work designs a new crossover-based genetic algorithm that uses mutation with a higher-than-usual mutation probability to increase the exploration speed and crossover with the parent to repair losses incurred by the more aggressive mutation. Expand
Black-Box Search by Unbiased Variation
TLDR
This paper introduces a more restricted black-box model for optimisation of pseudo-Boolean functions which it is claimed captures the working principles of many randomised search heuristics including simulated annealing, evolutionary algorithms, randomised local search, and others. Expand
The (1+1) Elitist Black-Box Complexity of LeadingOnes
TLDR
The permutation- and bit-invariant version of LeadingOnes is regarded and it is proved that its (1+1) elitist black-box complexity is Ω(n2), a bound that is matched by (1-1)-type evolutionary algorithms. Expand
Introducing Elitist Black-Box Models: When Does Elitist Behavior Weaken the Performance of Evolutionary Algorithms?
TLDR
The elitist model is added, in which algorithms are required to base all decisions solely on (the relative performance of) a fixed number of the best search points sampled so far, and the concept of p-Monte Carlo black-box complexity is introduced, which measures the time it takes to optimize a problem with failure probability at most p. Expand
The $$(1+1)$$(1+1) Elitist Black-Box Complexity of LeadingOnes
TLDR
The permutation- and bit-invariant version of LeadingOnes is regarded and it is proved that its(1+1) elitist black-box complexity is VarOmega (n^2)Ω(n2), a bound that is matched by(1-1)-type evolutionary algorithms, a bound which shows that for LeadingOns the memory-restriction, together with the selection requirement, has a substantial impact on the best possible performance. Expand
Elitist Black-Box Models: Analyzing the Impact of Elitist Selection on the Performance of Evolutionary Algorithms
TLDR
This work proposes a new elitist black-box model, in which algorithms are required to base all decisions solely on (a fixed number of) the best search points sampled so far, and introduces the concept of $p-Monte Carlo black- box complexity, which measures the time it takes to optimize a problem with failure probability at most p. Expand
Optimal Parameter Choices via Precise Black-Box Analysis
TLDR
It is proved that the unary unbiased black-box complexity of the classic OneMax function class is n ln( n) -- cn ± o(n) for a constant c between 0.2539 and 0.2665. Expand
Black-box Complexity of Parallel Search with Distributed Populations
TLDR
The results confirm that rings and torus graphs can lead to higher black-box complexities, compared to unrestricted communication, and determine cut-off points for the number of populations λ, above which no distributed black- box algorithm can achieve linear speedups. Expand
Faster black-box algorithms through higher arity operators
TLDR
The work of Lehre and Witt (GECCO 2010) on the unbiased black- box model is extended by considering higher arity variation operators by showing that already for binary operators the black-box complexity of LeadingOnes drops from Θ(<i>n</i><sup>2</sup>) for unary operators to <i>O</i>(<i-i> log <i*n> log<i+i>) in the binary case. Expand
...
1
2
3
4
5
...