Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]
Continuous Hierarchical Representations with Poincaré Variational Auto-Encoders
Authors: Emile Mathieu, Charline Le Lan, Chris J. Maddison, Ryota Tomioka, Yee Whye Teh
NeurIPS 2019 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically show better generalisation to unseen data than the Euclidean counterpart, and can qualitatively and quantitatively better recover hierarchical structures. (Abstract) ... We implemented our model and ran our experiments within the automatic differentiation framework Py Torch (Paszke et al., 2017). ... 5 Experiments |
| Researcher Affiliation | Collaboration | Emile Mathieu EMAIL, Department of Statistics, University of Oxford, United Kingdom; Chris J. Maddison EMAIL, Deep Mind, London, United Kingdom; Ryota Tomioka EMAIL, Microsoft Research, Cambridge, United Kingdom |
| Pseudocode | Yes | Algorithm 1 Hyperbolic normal sampling scheme |
| Open Source Code | Yes | We open-source our code for reproducibility and to benefit the community 1. 1https://github.com/emilemathieu/pvae |
| Open Datasets | Yes | We assess our modelling assumption on data generated from a branching diffusion process... The MNIST (Le Cun and Cortes, 2010) dataset... We demonstrate these capabilities on three network datasets: a graph of Ph.D. advisor-advisee relationships (Nooy et al., 2011), a phylogenetic tree expressing genetic heritage (Hofbauer et al., 2016; Sanderson and Eriksson, 1994) and a biological set representing disease relationships (Goh et al., 2007; Rossi and Ahmed, 2015). |
| Dataset Splits | No | The paper mentions 'Learning rates are chosen by cross-validation among {0.01, 0.005, 0.001}' in Appendix C, implying a validation process for hyperparameter tuning. However, it does not explicitly provide specific percentages or sample counts for a dedicated validation dataset split, nor does it explicitly mention a 'validation set' in the main text with its dimensions. |
| Hardware Specification | No | The paper states, 'We implemented our model and ran our experiments within the automatic differentiation framework Py Torch (Paszke et al., 2017).', but it does not specify any details about the hardware used, such as CPU or GPU models, memory, or specific computing environments. |
| Software Dependencies | No | The paper mentions using 'Py Torch (Paszke et al., 2017)' and the 'Adam optimizer (Kingma and Ba, 2016)'. However, it does not provide specific version numbers for these software components (e.g., 'PyTorch 1.9' or 'Adam version X'), which are necessary for full reproducibility. |
| Experiment Setup | Yes | Experimental details are fully described in Appendix C. ... We train all models for 100 epochs using the Adam optimizer (Kingma and Ba, 2016) with a batch size of 100 for the synthetic data and 500 for MNIST. For the graph datasets we use a batch size of 1 for the node features and 200 for the graph edges. Learning rates are chosen by cross-validation among {0.01, 0.005, 0.001} and the prior dispersion for the P-VAE among {0.1, 0.3, 0.8, 1.0, 1.2, 1.4}. |