Flat Metric Minimization with Applications in Generative Modeling

Authors: Thomas Möllenhoff, Daniel Cremers

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In experiments, we show that the proposed shift to k > 0 leads to interpretable and disentangled latent representations which behave equivariantly to the specified oriented tangent planes.
Researcher Affiliation Academia 1Department of Informatics, Technical University of Munich, Garching, Germany.
Pseudocode No The paper describes mathematical formulations and implementation details, but does not include a clearly labeled pseudocode or algorithm block.
Open Source Code Yes See https://github.com/moellenh/flatgan for a PyTorch implementation to reproduce Fig. 6 and Fig. 7.
Open Datasets Yes For MNIST, we compute the tangent vectors manually by rotation and dilation of the digits, similar as done by Simard et al. (1992; 1998). For the small NORB example, the tangent vectors are given as differences between the corresponding images. As observed in the figures, the proposed formulation leads to interpretable latent codes.
Dataset Splits No The paper mentions using datasets like MNIST, small NORB, and tinyvideos but does not provide specific details on training, validation, or test splits (e.g., percentages, sample counts, or explicit cross-validation setup).
Hardware Specification No The paper does not provide any specific hardware details such as CPU/GPU models, memory, or cloud instance types used for running experiments.
Software Dependencies No The paper mentions 'PyTorch implementation' but does not specify version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes The specific hyperparameters, architectures and tangent vector setups used in practice3 are detailed in Appendix B.