Completely random measures for modelling block-structured sparse networks
Authors: Tue Herlau, Mikkel N. Schmidt, Morten Mørup
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The proposed method was evaluated on 11 network datasets (a description of how the datasets were obtained and prepared can be found in the supplementary material) using K = 200 in the truncated stick-breaking representation. As a criteria of evaluation we choose AUC score on held-out edges, i.e. predicting the presence or absence of unobserved edges using the imputation method described in the previous section. |
| Researcher Affiliation | Academia | DTU Compute Technical University of Denmark Richard Petersens plads 31, 2800 Lyngby, Denmark {tuhe,mns,mmor}@dtu.dk |
| Pseudocode | No | The paper describes inference steps in paragraph form, but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | 1Code available at http://people.compute.dtu.dk/tuhe/crmsbm. |
| Open Datasets | No | The paper mentions '11 network datasets (a description of how the datasets were obtained and prepared can be found in the supplementary material)' but does not provide specific access information (links, DOIs, formal citations) in the main text. |
| Dataset Splits | No | The paper mentions 'A fraction of 5% of the edges were removed and considered as held-out data' for evaluation, which serves as a test set, but does not explicitly specify comprehensive training, validation, and test splits. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory) were mentioned for the experimental setup. |
| Software Dependencies | No | No specific software dependencies with version numbers were mentioned. |
| Experiment Setup | Yes | The other parameters were fixed at α = 20K, τ = 1, σ = 0.5 and λa = λb = 1. ... All methods were evaluated for T = 2 000 iterations, and the latter half of the chains was used for link prediction. ... For the priors we selected uniform priors for σ, τ, α and a Gamma(2, 1) prior for β0, λa, λb. ... we applied 50 updates of Φℓfor each update of (zi)i and Aij. |