Hyperbolic Entailment Cones for Learning Hierarchical Embeddings
Authors: Octavian Ganea, Gary Becigneul, Thomas Hofmann
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5. Experiments. We evaluate the representational and generalization power of hyperbolic entailment cones and of other baselines using data that exhibits a latent hierarchical structure. We follow previous work (Nickel & Kiela, 2017; Vendrov et al., 2015) and use the full transitive closure of the Word Net noun hierarchy (Miller et al., 1990). Our binary classification task is link prediction for unseen edges in this directed acyclic graph. |
| Researcher Affiliation | Academia | 1Department of Computer Science, ETH Zurich, Switzerland. |
| Pseudocode | No | The paper describes an 'efficient algorithm for learning hierarchical embeddings' in text, but does not provide it in a structured pseudocode or algorithm block. |
| Open Source Code | Yes | Our code is publicly available9. 9https://github.com/dalab/hyperbolic_cones. |
| Open Datasets | Yes | We follow previous work (Nickel & Kiela, 2017; Vendrov et al., 2015) and use the full transitive closure of the Word Net noun hierarchy (Miller et al., 1990). |
| Dataset Splits | Yes | We split it into train validation test sets as follows. We first compute the transitive reduction7 of this directed acyclic graph... The remaining non-basic edges (578,477) are split into validation (5%), test (5%) and train (fraction of the rest). |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., CPU, GPU models, memory, or specific cloud instances) used for running the experiments. |
| Software Dependencies | No | The paper mentions 'stochastic gradient descent' and uses 'PyTorch' as an optimization framework in Appendix H (Training details), but it does not specify version numbers for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | Training details. We train our models with stochastic gradient descent (SGD) with a batch size of 200 and a constant learning rate of 1. We optimize hyperbolic cones (both our own and Poincar e Emb) with Riemannian SGD implemented in PyTorch. |