Per-Run Algorithm Selection with Warm-starting using Trajectory-based Features - Data
Zenodo (CERN European Organization for Nuclear Research)
Abstract
This repository contains the code and data for reproducing the results in the paper 'Per-Run Algorithm Selection with Warm-starting using Trajectory-based Features' The file 'figure_generation.ipnb' contains the code used to create the heatmaps shown in the paper. The file 'Raw_data' contains the raw performance data, for both running A1 and recording all relevant features and for running the switching algorithms for each of the 6 algorithms in the portfolio while recording only performance. The processed version of the A2 data into relative performance is available in 'perf_relative'. The file 'data-collection' contains the code used to run the switching algorithms with the used warmstarting procedures. The file 'compute_features.ipnyb' contains the code used to extract the timeseries features, and those features themselves are included in 'TS_Features'. The file 'AlgorithmSelector' contains the scripts and data used to create the algorithm selector. The files 'selector_5_as_f' and 'selector_ng_f' contain the results of running the final models on the test-instances and nevergrad functions respectively.
Authors 7
-
Centre National de la Recherche Scientifique · Sorbonne Université
Affiliation as printed
Sorbonne Université, CNRS
-
Jožef Stefan Institute · Jožef Stefan International Postgraduate School
Affiliation as printed
Jožef Stefan Institute
-
Affiliation as printed
Leiden Institute for Advanced Computer Science
-
Affiliation as printed
Leiden Institute for Advanced Computer Science
-
Affiliation as printed
Leiden Institute for Advanced Computer Science
-
Jožef Stefan Institute · Jožef Stefan International Postgraduate School
Affiliation as printed
Jožef Stefan Institute
-
Centre National de la Recherche Scientifique · Sorbonne Université
Affiliation as printed
Sorbonne Université, CNRS