Automated Algorithm Selection: Survey and Perspectives

Abstract

It has long been observed that for practically any computational problem that has been intensely studied, different instances are best solved using different algorithms. This is particularly pronounced for computationally hard problems, where in most cases, no single algorithm defines the state of the art; instead, there is a set of algorithms with complementary strengths. This performance complementarity can be exploited in various ways, one of which is based on the idea of selecting, from a set of given algorithms, for each problem instance to be solved the one expected to perform best. The task of automatically selecting an algorithm from a given set is known as the per-instance algorithm selection problem and has been intensely studied over the past 15 years, leading to major improvements in the state of the art in solving a growing number of discrete combinatorial problems, including propositional satisfiability and AI planning. Perinstance algorithm selection also shows much promise for boosting performance in solving continuous and mixed discrete/continuous optimisation problems.This survey provides an overview of research in automated algorithm selection, ranging from early and seminal works to recent and promising application areas. Different from earlier work, it covers applications to discrete and continuous problems, and discusses algorithm selection in context with conceptually related approaches, such as algorithm configuration, scheduling or portfolio selection. Since informative and cheaply computable problem instance features provide the basis for effective perinstance algorithm selection systems, we also provide an overview of such features for discrete and continuous problems. Finally, we provide perspectives on future work in the area and discuss a number of open research challenges.

Benchmarking discrete optimization heuristics with IOHprofiler

“…Another interesting extension of IO-Hanalyzer would be the design of modules that allow us to couple the performance evaluation with an analysis of the fitness landscape of the considered problems. Such feature-based analyses are at the heart of algorithm selection techniques [30], which use landscape features and performance data to build a model that predicts how the tested algorithms will perform on a previously not tested problem. Similar approaches can be found in per-instance-algorithm configuration (PIAC) approaches, which have recently shown very promising performance in the context of continuous black-box optimization [6].…”

Predicting Good Configurations for GitHub and Stack Overflow Topic Models

“…As algorithm selection is often implemented using machine learning [28], [29], we need two preparation steps: (i) instance features that characterise instances numerically, (ii) performance data of each algorithm on each instance. We have already characterised our corpora in Section III-B, so we only need to run each of the 17 configurations on all corpora.…”

Instruction-level design of local optimisers using push GP

“…Given that there are many optimisation algorithms available, and that each algorithm has many variants and potential hybrids, this process of selecting the optimal optimiser has the capacity to be very involved and time consuming. One way of addressing this is to use a machine learning algorithm to select or design an optimiser on the user's behalf [Rice, 1976, Kerschke et al, 2019. In particular, the hyperheuristics [Burke et al, 2013, Swan et al, 2018 community has been exploring this idea for some time, typically by using evolutionary algorithms to select or generate the heuristics used by a particular metaheuristic framework, or generating new metaheuristic frameworks by combining existing heuristics.…”