Parallelized Hitting Set Computation for Model-Based Diagnosis

Authors: Dietmar Jannach, Thomas Schmitz, Kostyantyn Shchekotykhin

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluations using different benchmark problems show that parallelization can help to significantly reduce the required running times. Additional simulation experiments were performed to understand how the characteristics of the underlying problem structure impact the achieved performance gains.
Researcher Affiliation Academia Dietmar Jannacha, Thomas Schmitza, Kostyantyn Shchekotykhinb a TU Dortmund, Germany b Alpen-Adria Universit at Klagenfurt, Austria {firstname.lastname}@tu-dortmund.de, kostya@ifit.uni-klu.ac.at
Pseudocode Yes Algorithm 1: DIAGNOSE: Main loop. Algorithm 2: EXPAND: Node expansion logic. Algorithm 3: CHECKANDADDPATH: Create a path label. Algorithm 4: Level-wise parallelization approach. Algorithm 5: Fully parallelized HS-tree construction. Algorithm 6: Node expansion of the full parallelization.
Open Source Code No The paper does not provide any links to open-source code for its methodology, nor does it state that the code is available.
Open Datasets Yes For these experiments, we selected the first five systems of the DX Competition 2011 Synthetic Track1 (see Table 1). ... we used a number of CSP instances from the 2008 CP solver competition4.
Dataset Splits No The paper describes its experimental procedure and uses different datasets but does not specify train/validation/test splits explicitly. It talks about '20 scenarios with injected faults' and 'repeated all tests 100 times' but not how the data itself was split into train/validation/test sets for model training or evaluation.
Hardware Specification Yes We used a standard laptop computer (Intel i7-3632QM, 4 cores with Hyper-Threading, 16GB RAM) running Windows 8.
Software Dependencies Yes The diagnosis algorithms were implemented in Java using Choco 2 as a constraint solver and QUICKXPLAIN for conflict detection.
Experiment Setup Yes We limited the search depth to 4 for all experiments to speed up the benchmark process.