Decomposing Parameter Estimation Problems
Authors: Khaled S Refaat, Arthur Choi, Adnan Darwiche
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments are structured as follows. Given a Bayesian network G, we generate a dataset D while ensuring that a certain percentage of variables are observed, with all others hidden. Using dataset D, we estimate the parameters of network G using two methods. The first uses the classical EM on network G and dataset D. |
| Researcher Affiliation | Academia | Khaled S. Refaat, Arthur Choi, Adnan Darwiche Computer Science Department University of California, Los Angeles {krefaat,aychoi,darwiche}@cs.ucla.edu |
| Pseudocode | No | The paper outlines steps for the decomposition technique (e.g., "1. Identify the components... 2. Construct a sub-network..."), but these are presented as numbered text descriptions rather than a formal pseudocode block or algorithm structure. |
| Open Source Code | No | The paper mentions and compares against implementations in SMILE and SAMIAM, providing their URLs, but does not state that the code for their own D-EM method is open-source or available. |
| Open Datasets | Yes | For example, in the UCI repository: the internet advertisements dataset has 1558 variables, only 3 of which have missing values; the automobile dataset has 26 variables, where 7 have missing values; the dermatology dataset has 34 variables, where only age can be missing; and the mushroom dataset has 22 variables, where only one variable has missing values [1]. |
| Dataset Splits | No | The paper mentions generating datasets and using a specific number of examples per dataset (e.g., "210 examples per dataset"), but it does not specify how these datasets were split into training, validation, or test sets, nor does it refer to predefined splits. |
| Hardware Specification | No | The paper does not specify the hardware (e.g., CPU, GPU, memory, specific server configurations) used to run the experiments. |
| Software Dependencies | No | The paper mentions using SMILE's C++ API and SAMIAM's Java API, and the Hugin algorithm as the underlying inference engine, but it does not provide specific version numbers for any of these software components. |
| Experiment Setup | Yes | In our experiments, we used a convergence test based on parameter change. In particular, when the absolute change in every parameter falls below the set threshold of 10 4, convergence is declared by EM. ... In our experiments, we used a threshold of 10 5 for D-EM, which is tighter than the threshold used for EM. |