Consistent Nonparametric Methods for Network Assisted Covariate Estimation
Authors: Xueyu Mao, Deepayan Chakrabarti, Purnamrita Sarkar
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4. Experiments We evaluate the accuracy and speed of CN-VEC and SVD-RBF on several simulated and real-world networks. |
| Researcher Affiliation | Academia | 1Department of Computer Science. 2Department of Information, Risk, and Operations Management. 3Department of Statistics and Data Sciences. The University of Texas at Austin, TX, USA. |
| Pseudocode | Yes | Algorithm 1 CN-VEC: model-agnostic algorithm" and "Algorithm 2 SVD-RBF: nonparametric regression for low rank models with the RBF kernel Kθ (v1, v2) |
| Open Source Code | No | The paper references third-party code for `node2vec` and `NOBE` but does not provide a link or explicit statement for its own methodology's source code. |
| Open Datasets | Yes | We evaluated our method on two citation networks, namely Cora (Mc Callum et al., 2000) and Cite Seer (Giles et al., 1998)3, and one social network, namely Sinanet (Jia et al., 2017)4. |
| Dataset Splits | No | For each network, we vary the fraction of nodes with unknown covariates from 0.5 to 0.9. For each fraction, we randomly select the nodes with unknown covariates and predict their covariates using the various algorithms. |
| Hardware Specification | Yes | All experiments are performed with Matlab R2018b on servers with 24-core Intel Xeon X5675 and 99GB RAM. |
| Software Dependencies | Yes | All experiments are performed with Matlab R2018b |
| Experiment Setup | Yes | For each network, we vary the fraction of nodes with unknown covariates from 0.5 to 0.9... Each network has n = 2, 500 nodes and latent dimension d = 5 by default. The node covariates are generated by Xi = βT zi+N(0, .1)... Then, given a node i, it picks the top-10 most similar nodes according to the W, and calculates the weighted average of their node covariates, with W as the weights. |