Disentangling Disentanglement in Variational Autoencoders
Authors: Emile Mathieu, Tom Rainforth, N Siddharth, Yee Whye Teh
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We confirm these results empirically, while further using our decomposition framework to show that simple manipulations to the prior can improve disentanglement, and other decompositions, with little or no detriment to the reconstruction accuracy. Further, motivated by our analysis, we propose an alternative objective that takes into account the distinct needs of the two factors of decomposition, and use it to learn clustered and sparse representations as demonstrations of alternative forms of decomposition. |
| Researcher Affiliation | Academia | Emile Mathieu * 1 Tom Rainforth * 1 N. Siddharth * 2 Yee Whye Teh 1 [...] 1Department of Statistics 2Department of Engineering, University of Oxford. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | An implementation of our experiments and suggested methods is provided at http://github.com/iffsid/disentangling-disentanglement. |
| Open Datasets | Yes | 2D Shapes dataset (Matthey et al., 2017), pinwheels dataset from (Johnson et al., 2016) and Fashion-MNIST dataset (Xiao et al., 2017) |
| Dataset Splits | No | The paper mentions the datasets used for experiments (2D Shapes, pinwheels, Fashion-MNIST) but does not provide specific details on how these datasets were split into training, validation, and test sets. |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependency versions (e.g., library names with version numbers) needed to replicate the experiments. |
| Experiment Setup | No | The paper refers to 'full experimental details' being in Appendix B, but the main text does not provide specific hyperparameter values or concrete training configurations such as learning rates, batch sizes, or optimizer settings. |