Demystifying Inductive Biases for (Beta-)VAE Based Architectures

Authors: Dominik Zietlow, Michal Rolinek, Georg Martius

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate a set of approaches on slightly modified versions of the two leading datasets in which each image undergoes a modification inducing little variance. We report drastic drops of disentanglement performance on the altered datasets. The resulting MIG scores are listed in Tab. 1, other metrics are listed in the supplementary materials.
Researcher Affiliation Academia 1Max Planck Institute for Intelligent Systems, T ubingen, Germany.
Pseudocode No No pseudocode or clearly labeled algorithm block was found in the paper.
Open Source Code No The paper does not provide a specific link or explicit statement about the release of its own source code for the described methodology. It mentions using the codebase from Locatello et al. (2019).
Open Datasets Yes We close this gap for VAE-based architectures on the two most common datasets, namely d Sprites (Matthey et al., 2017) and Shapes3d (Burgess & Kim, 2018).
Dataset Splits No The paper does not explicitly state specific training, validation, or test dataset splits (percentages or counts) used for reproducibility. It refers to using existing datasets and hyperparameters from a disentanglement library.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper mentions that 'Our experiments are built on their codebase (Locatello et al., 2019)' but does not provide specific software dependencies or version numbers for reproducibility.
Experiment Setup No The paper generally refers to using 'regularization strengths as reported in the literature (or better tuned values)' and 'other hyperparameters are taken from the disentanglement library (Locatello et al., 2019)' without providing concrete hyperparameter values or detailed training configurations.