Finding Good Partial Assignments during Restart-Based Branch and Bound Search

Authors: Hongbo Li, Jimmy H.M. Lee

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on the Mini Zinc benchmark suite show how our approach brings significant improvements to a blackbox COP solver equipped with the state of the art search techniques.
Researcher Affiliation Academia Hongbo Li1 and Jimmy H.M. Lee2 1School of Information Science and Technology, Northeast Normal University, Changchun, China 2Department of Computer Science and Engineering, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong lihb905@nenu.edu.cn, jlee@cse.cuhk.edu.hk
Pseudocode Yes Algorithm 1: Restart-based depth-first BBS
Open Source Code Yes The source code of RGPA implemented in Choco 4.10.8 is available at https://github.com/lihb905/rgpa.
Open Datasets Yes The experiments were run with the Mini Zinc benchmark suite from https://github.com/MiniZinc/minizinc-benchmarks.
Dataset Splits No The paper states it uses the Mini Zinc benchmark suite but does not provide specific details on dataset splits (e.g., percentages, sample counts, or citations to predefined splits) for training, validation, or testing.
Hardware Specification Yes The environment is JDK8 under Cent OS 6.4 with Intel Xeon CPU E7-4820@2.00GHz processor and 58 GB RAM.
Software Dependencies Yes The environment is JDK8 under Cent OS 6.4... The experiments were run in Choco solver (version 4.10.8)
Experiment Setup Yes The timeout is set to 2 hours. We have used a unique random seed 0 throughout the experiments... The Luby sequence... based restart strategy which achieves a good balance between frequent and extended restarts is employed. The initial restart cutoff is set to n, the number of variables. The default nogood recording (Lecoutre et al. 2007) technique in Choco is also used.