Zero-Suppressed Sentential Decision Diagrams

Authors: Masaaki Nishino, Norihito Yasuda, Shin-ichi Minato, Masaaki Nagata

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that ZSDDs are smaller than SDDs or ZDDs for a standard benchmark dataset. We compared the sizes of ZSDDs, SDDs, and ZSDDs that used right-linear vtrees. As the dataset, we used LGSynth89 benchmark dataset.
Researcher Affiliation Collaboration Masaaki Nishino1 and Norihito Yasuda2 and Shin-ichi Minato2 and Masaaki Nagata1 1NTT Communication Science Laboratories, NTT Corporation 2Graduate School of Information Science and Technology, Hokkaido University
Pseudocode Yes Algorithm 1 Apply(α, β, ). α, β are compressed, lightlytrimmed, and normalized ZSDDs, and is a binary operation for families of sets. Algorithm 2 Change(α, X)
Open Source Code Yes 1Our sample software of ZSDD is available at https://github. com/nsnmsak/zsdd/.
Open Datasets Yes As the dataset, we used LGSynth89 benchmark dataset.
Dataset Splits No No specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, or testing was provided. The paper uses the LGSynth89 benchmark dataset, which consists of CNF instances, for evaluating the size of decision diagrams rather than training a model.
Hardware Specification No No specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running experiments were mentioned.
Software Dependencies No The paper mentions using 'SDD package' and 'Cu DD' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes The vtrees for ZSDDs and SDDs were determined by the dynamic reordering algorithm (Choi and Darwiche 2013) implemented in the SDD package. We used a balanced vtree as the initial vtree for reordering.