Procrastinating with Confidence: Near-Optimal, Anytime, Adaptive Algorithm Configuration

Authors: Robert Kleinberg, Kevin Leyton-Brown, Brendan Lucier, Devon Graham

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show empirically both that such settings arise frequently in practice and that the anytime property is useful for finding good configurations quickly. 5 Experimental Results We experiment with SPC on the benchmark set of runtimes generated by Weisz et al. (2018b) for testing LEAPSANDBOUNDS.
Researcher Affiliation Collaboration Robert Kleinberg Department of Computer Science Cornell University rdk@cs.cornell.edu Kevin Leyton-Brown Department of Computer Science University of British Columbia kevinlb@cs.ubc.ca Brendan Lucier Microsoft Research brlucier@microsoft.com Devon Graham Department of Computer Science University of British Columbia drgraham@cs.ubc.ca
Pseudocode Yes Algorithm 1: Structured Procrastination w/ Confidence
Open Source Code Yes 3Code to reproduce experiments is available at https://github.com/drgrhm/alg_config
Open Datasets Yes We experiment with SPC on the benchmark set of runtimes generated by Weisz et al. (2018b) for testing LEAPSANDBOUNDS. This data consists of pre-computed runtimes for 972 configurations of the minisat (Sorensson & Een, 2005) SAT solver on 20118 SAT instances generated using CNFuzz DD4.4http://fmv.jku.at/cnfuzzdd/
Dataset Splits No The paper uses a benchmark set of pre-computed runtimes but does not specify any explicit training, validation, or test dataset splits.
Hardware Specification No The paper mentions 'CPU time in days' for experimental runtime but does not provide specific hardware details such as CPU/GPU models, memory, or other system specifications.
Software Dependencies No The paper mentions the 'minisat' SAT solver used to generate the dataset but does not list specific software dependencies with version numbers required to replicate the experiments.
Experiment Setup No The paper describes the benchmark data and comparisons made, but does not provide specific hyperparameters or system-level training settings for SPC within its experimental setup.