Property Controllable Variational Autoencoder via Invertible Mutual Dependence
Authors: Xiaojie Guo, Yuanqi Du, Liang Zhao
ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Quantitative and qualitative evaluations confirm that the PCVAE outperforms the existing models by up to 28% in capturing and 65% in manipulating the desired properties. |
| Researcher Affiliation | Academia | Xiaojie Guo Department of IST George Mason University Fairfax, VA 22030, USA xguo7@gmu.edu Yuanqi Du Department of CS George Mason University Fairfax, VA 22030, USA ydu6@gmu.edu Department of CS Emory University Atlanta, GA 30322, USA liang.zhao@emory.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code for the proposed PCVAE is available at:https://github.com/xguo7/PCVAE. |
| Open Datasets | Yes | The d Sprites dataset (Matthey et al., 2017) consists of 2D shapes procedurally generated from ground truth independent semantic factors. The 3Dshapes dataset (Burgess & Kim, 2018) consists of 3D shapes procedurally generated from ground truth independent semantic factors. The QM9 dataset (Ramakrishnan et al., 2014) consists of 134k stable small organic molecules |
| Dataset Splits | No | For the d Sprites, 3Dshapes, and QM9 datasets, the paper specifies "training/testing set split" but does not explicitly mention a separate validation set split or its size/methodology. |
| Hardware Specification | Yes | All experiments were conducted on a 64-bit machine with an NVIDIA GPU (GTX 1080 Ti, 11016 MHz, 11 GB GDDR5). |
| Software Dependencies | No | The paper does not provide specific software names with version numbers for dependencies, only general mentions of MLPs, CNNs, GNNs, and spectral normalization. |
| Experiment Setup | Yes | The architectures and hyper-parameters can be found in Appendix B. (In Appendix B, Table 6 specifies 'Learning rate', 'Batch size', 'α', 'ρ', 'Num iteration', 'c (spectral norm)' for d Sprites and QM9 datasets). |