Optimality of variational inference for stochasticblock model with missing links

Authors: Solenne Gaucher, Olga Klopp

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We complement our results with numerical studies of simulated and real networks, which confirm the advantages of this estimator over current methods. (...) Finally, in Section 4 we provide an extensive numerical study both on synthetic and real-life data which shows clear advantages of our estimator over current methods.
Researcher Affiliation Academia Solenne Gaucher D epartement de Math ematiques d Orsay, Universit e Paris-Saclay, Orsay, France; Olga Klopp ESSEC Business School, Cergy, France CREST, ENSAE, Palaiseau, France
Pseudocode No The paper mentions that "The expectation maximization (EM) algorithm derived in [48] can be used to iteratively compute the variational estimator. For details, see Appendix C." However, the provided text does not contain a clearly labeled pseudocode or algorithm block within its main sections, and without access to Appendix C, it cannot be confirmed if this paper explicitly presents the pseudocode itself.
Open Source Code No The paper mentions external packages such as "miss SBM" and "soft Impute" that implement certain methods, but it does not state that the authors are providing open-source code for the methodology described in this paper.
Open Datasets Yes We apply our algorithm to analyze a network of interactions within a French elementary school collected by the authors of [47]. (...) network of co-authorship between scientists working on network analysis, first analysed in [43].
Dataset Splits Yes We use the observations collected on Day 1 estimate the matrix , and evaluate those estimators on the network of interactions corresponding to Day 2. (...) We introduce 50% of missing values in the dataset. We train the three estimators on the observed entries of the adjacency matrix, and we use the unobserved entries to evaluate their imputation error.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud instances) used for running its experiments.
Software Dependencies No The paper mentions software packages like "miss SBM" and "soft Impute" but does not specify their version numbers or other ancillary software with version information required for reproducibility.
Experiment Setup No The paper describes general parameters for synthetic data generation (e.g., "three-communities stochastic block model," "50% missing values") and how real-world data was preprocessed (e.g., interaction duration). However, it does not provide specific hyperparameters for the variational inference estimator itself (e.g., learning rates, number of epochs/iterations if applicable), nor detailed system-level training settings for their proposed method.