Multi-Pass High-Level Presolving
Authors: Kevin Leo, Guido Tack
IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show that this approach can lead to both faster compilation and more efficient solver-level programs. ... The experiments summarised in this section measure the impact of the new approach on compile-time overhead, program size and solving performance. We present results of two pass compilation where Gecode is used in the first pass. The experiments used machines with dual 2.00GHz Intel Quad Core Xeon E5405 processors with 16GB of RAM. 300 instances from the previous three years of Mini Zinc Challenge problems were selected, covering 49 Mini Zinc models. |
| Researcher Affiliation | Academia | Kevin Leo and Guido Tack Faculty of IT, Monash University, Australia and National ICT Australia (NICTA) Victoria {kevin.leo,guido.tack}@monash.edu |
| Pseudocode | No | The paper includes code listings and figures that show examples of Mini Zinc and Flat Zinc code, but it does not contain any structured pseudocode or algorithm blocks describing the proposed multi-pass presolving process. |
| Open Source Code | No | The paper mentions and cites the Mini Zinc 2.0 compiler and its availability at minizinc.org, which is a tool used in the research, but it does not provide access to the specific source code written by the authors for the multi-pass presolving methodology described in the paper. |
| Open Datasets | No | The paper states, '300 instances from the previous three years of Mini Zinc Challenge problems were selected, covering 49 Mini Zinc models.' While Mini Zinc Challenge problems are generally known, the paper does not provide a specific URL, DOI, or formal citation to the exact dataset instances used in their experiments, or clearly indicate their public availability beyond this general mention. |
| Dataset Splits | No | The paper describes using '300 instances' for experiments but does not provide any details on how these instances were split into training, validation, or test sets. The analysis focuses on performance comparisons (presolved vs. non-presolved) rather than model training evaluation splits. |
| Hardware Specification | Yes | The experiments used machines with dual 2.00GHz Intel Quad Core Xeon E5405 processors with 16GB of RAM. |
| Software Dependencies | Yes | The experiments used the Mini Zinc 2.0 compiler... We used CPLEX version 12.6 for the MIP experiments. |
| Experiment Setup | Yes | Compilation and solving both had an 8Gb memory limit. ... To control for the erratic behaviour of heuristics in MIP solvers [Fischetti and Monaci, 2014] we ran each instance six times with different random seeds. |