WatE: A Wasserstein t-distributed Embedding Method for Information-enriched Graph Visualization
Authors: Minjie Cheng, Dixin Luo, Hongteng Xu
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Both subjective and objective evaluations demonstrate that Wat E achieves encouraging performance in various graph visualization and clustering tasks. To demonstrate the effectiveness of Wat E, we test it on various graph datasets and evaluate the graph visualization results on both subjective visual effects and objective clustering accuracy. We apply four TU graph datasets (Morris et al. 2020), including IMDB-B (social networks), MUTAG, PTC-MR (molecular datasets) and AIDS (biomedical graph datasets). The quantitative clustering results of different methods are shown in Table 1. In Figure 5(a), when 0.1 α 10 and 1 β 100, the performance of Wat E is stable, demonstrating its robustness to hyperparameters. In Figure 5(b), when β = 0, we learn the model without the covariance regularizer. |
| Researcher Affiliation | Academia | Minjie Cheng1, Dixin Luo2, Hongteng Xu1,3* 1Gaoling School of Artifcial Intelligence, Renmin University of China 2School of Computer Science and Technology, Beijing Institute of Technology 3Beijing Key Laboratory of Big Data Management and Analysis Methods EMAIL |
| Pseudocode | Yes | Algorithm 1: The Rt-SNE framework for Wat E model |
| Open Source Code | Yes | Code https://github.com/minjiecheng/Wat E. The algorithmic scheme is shown in Algorithm 1, and Newton-Schulz algorithm of bΣ 1 2n and conditional gradient algorithm of GW and FGW distance are presented at https://github.com/minjiecheng/Wat E. |
| Open Datasets | Yes | We apply four TU graph datasets (Morris et al. 2020), including IMDB-B (social networks), MUTAG, PTC-MR (molecular datasets) and AIDS (biomedical graph datasets). |
| Dataset Splits | Yes | For each dataset, we train our model on 80% of the graphs and then visualize the entire dataset. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment. |
| Experiment Setup | Yes | We set α = 1 and β = 100 by default. In Figure 5(a), when 0.1 α 10 and 1 β 100, the performance of Wat E is stable, demonstrating its robustness to hyperparameters. Applying stochastic gradient descent, the t-SNE method solves (1) and visualizes the data by the learned embeddings. Update θ via Adam (Kingma and Ba 2015). Following the t-SNE method (Van der Maaten and Hinton 2008), we determine this bandwidth by the bisection search, making the entropy of the conditional distribution {pn |n}N n =1 equal to a predefined perplexity. |