Search Trajectory Networks (STNs) are visualizations of directed graphs designed to analyze the behavior of stochastic optimization algorithms such as metaheuristics. Their purpose is to provide researchers with a too...
详细信息
ISBN:
(纸本)9783031629211;9783031629228
Search Trajectory Networks (STNs) are visualizations of directed graphs designed to analyze the behavior of stochastic optimization algorithms such as metaheuristics. Their purpose is to provide researchers with a tool that allows them to gain a deeper understanding of the behavior exhibited by multiple algorithms when applied to a specific instance of an optimization problem. In this short paper, we present two examples of our work in which STN graphics have helped us to discover interesting and useful algorithm/problem characteristics.
In black-box optimization, it is essential to understand why an algorithm instance works on a set of problem instances while failing on others and provide explanations of its behavior. We propose a methodology for for...
详细信息
ISBN:
(纸本)9798400701191
In black-box optimization, it is essential to understand why an algorithm instance works on a set of problem instances while failing on others and provide explanations of its behavior. We propose a methodology for formulating an algorithm instance footprint that consists of a set of problem instances that are easy to be solved and a set of problem instances that are difficult to be solved, for an algorithm instance. This behavior of the algorithm instance is further linked to the landscape properties of the problem instances to provide explanations of which properties make some problem instances easy or challenging. The proposed methodology uses meta-representations that embed the landscape properties of the problem instances and the performance of the algorithm into the same vector space. These meta-representations are obtained by training a supervised machine learning regression model for algorithm performance prediction and applying model explainability techniques to assess the importance of the landscape features to the performance predictions. Next, deterministic clustering of the meta-representations demonstrates that using them captures algorithm performance across the space and detects regions of poor and good algorithm performance, together with an explanation of which landscape properties are leading to it.
This paper investigates reasons behind the behavior of constructive Solution-Guided Search (SGS) on job-shop scheduling optimization problems. In particular, two, not mutually exclusive, hypotheses are investigated: (...
详细信息
This paper investigates reasons behind the behavior of constructive Solution-Guided Search (SGS) on job-shop scheduling optimization problems. In particular, two, not mutually exclusive, hypotheses are investigated: (1) Like randomized restart, SGS exploits heavy-tailed distributions of search cost;and (2) Like local search, SGS exploits search space structure such as the clustering of high-quality solutions. Theoretical and experimental evidence strongly support both hypotheses. Unexpectedly, the experiments into the second hypothesis indicate that the performance of randomized restart and standard chronological backtracking are also correlated with search space structure. This result leaves open the question of finding the mechanism by which such structure is exploited as well as suggesting a deeper connection between the performance of constructive and local search.
This article demonstrates the advantages of using visualization as part of the modeling process. Several examples are given to show how visualization can help developers to more completely understand the range of beha...
详细信息
This paper presents a methodology for behavior characterization of an algorithm in terms of the parametric description of input images. To develop the work we have selected an algorithm which implements a model of tex...
详细信息
This paper presents a methodology for behavior characterization of an algorithm in terms of the parametric description of input images. To develop the work we have selected an algorithm which implements a model of texture perception and provides a texture representation. The approach is based on the definition of an input parametric texture space, where parameters are related to texton attributes. Multidimensional scaling provides a dimensional reduction of space of representation. It allows interpretation of the behavior of the algorithm in a low-dimensional space where points represent textures and distances represent dissimilarities between textures, preserving the metric of the algorithm representation in a monotonic sense. The resulting behavior space establishes the basis to construct a quantitative causal model of an algorithm.
暂无评论