Spectral Approximate Inference

Authors: Sejun Park, Eunho Yang, Se-Young Yun, Jinwoo Shin

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we evaluate our algorithms on diverse environments including both synthetic and UAI datasets to corroborate our theorem and claims.
Researcher Affiliation Collaboration 1School of Electrical Engineering, KAIST, Daejeon, Korea 2School of Computing, KAIST, Daejeon, Korea 3Graduate School of AI, KAIST, Daejeon, Korea 4AITRICS, Seoul, Korea 5Department of Industrial & System Engineering, KAIST, Daejeon, Korea.
Pseudocode Yes Algorithm 1 Spectral inference for low-rank GMs; Algorithm 2 Spectral inference for high-rank GMs
Open Source Code No The paper does not provide an explicit statement or a link to the source code for the methodology described.
Open Datasets Yes We also evaluate our algorithms with GMs on grid graphs in a dataset for UAI 2014 inference competition. It provides 8 GMs on grid graphs, where 4 of them are of 100 vertices (10x10) and the other 4 are of 400 vertices (20x20). Figure 3f reports the approximation error and the running time of each algorithm.
Dataset Splits No The paper uses synthetic and UAI datasets for evaluation but does not explicitly describe specific train/validation/test splits, sample counts, or mention cross-validation setups.
Hardware Specification No The paper mentions running algorithms using 'a single thread of CPU' but does not provide specific hardware details such as CPU model, GPU models, or memory specifications used for the experiments.
Software Dependencies No The paper states, 'For solving the semi-definite programming (SDP) (15), we use CVX (Grant et al., 2008) with SDPT3 solver (Toh et al., 1999) using MATLAB.' However, it does not provide specific version numbers for CVX, SDPT3, or MATLAB.
Experiment Setup Yes Throughout our all experiments, we fix c = p|λj|/1000 for Algorithm 1 and Algorithm 2 to bound its running time regardless of eigenvalues. For fair comparisons, we choose 200 iterations for BP, 1000 iterations for MF and 10 ibound for MBE and WMBE.