Learning Bayesian Networks with Incomplete Data by Augmentation

Authors: Tameem Adel, Cassio de Campos

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform a wide range of experiments to demonstrate the benefits of learning Bayesian networks with such new approach.
Researcher Affiliation Academia Tameem Adel University of Manchester, UK tameem.hesham@gmail.com Cassio P. de Campos Queen s University Belfast, UK c.decampos@qub.ac.uk
Pseudocode No The paper describes the algorithms in prose but does not provide structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions external tools and their code availability (e.g., 'Gobnilp...with the code available from https://www.cs.york.ac.uk/aig/sw/gobnilp/' and 'structural EM...implementation available at https://github.com/cassiopc/csda-dataimputation'), but it does not provide a link or statement about the open-sourcing of the authors' own implementation described in the paper.
Open Datasets Yes We perform experiments on simulated as well as real-world data. First, we employ the original Bayesian network model for Breast Cancer (Almeida et al. 2014)... Second, we use the Bayesian network that has been learned from the Prostate Cancer data... (Sarabando 2011; Almeida et al. 2014)... Third, the well-known ASIA network is used (Lauritzen and Speigelhalter 1988). The LUCAS dataset contains data of the LUCAS causal Bayesian network (Fogelman-Soulie 2008)... The Single Proton Emission Computed Tomography (SPECT) dataset consists of binary data denoting partial diagnosis from SPECT images (Lichman 2013). The dataset used in this experiment is taken from a smoking cessation study as described in Gruder et al. (1993). The Car Evaluation dataset (Blake and Merz 1998; Lichman 2013).
Dataset Splits Yes The evaluation metric is the accuracy of the test instances using a cross-validation approach, as usually done in classification experiments. ... (100-fold cross-validation in the MNAR setting case)...
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments.
Software Dependencies No The paper mentions software components like 'Gobnilp' and 'WEKA', but it does not specify their version numbers, which is required for reproducibility.
Experiment Setup Yes We use maximum number of parents, k = 3, and use t = 1.