Learning disentangled representations via product manifold projection

Authors: Marco Fumero, Luca Cosmo, Simone Melzi, Emanuele Rodola

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

Reproducibility Variable Result LLM Response
Research Type Experimental We widely test our approach on synthetic datasets and more challenging real-world scenarios, outperforming the state of the art in several cases.
Researcher Affiliation Academia 1Sapienza, University of Rome, Rome, Italy 2Universit a della Svizzera italiana, Lugano, Switzerland.
Pseudocode No The paper describes the model and losses textually and with diagrams, but does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about releasing source code for the described methodology, nor does it provide a direct link to a code repository.
Open Datasets Yes We adopted 4 widely used synthetic datasets in order to evaluate the effectivness of our method, namely DSprites (Higgins et al., 2017), Shapes3D (Kim & Mnih, 2018), Cars3D (Reed et al., 2015), Small NORB (Le Cun et al., 2004) . For the experiments on FAUST we used a Point Net (Qi et al., 2017) architecture and a simple MLP as a decoder.
Dataset Splits No The paper discusses training and testing, and evaluation metrics, but does not explicitly provide details about training/validation/test dataset splits (e.g., percentages, sample counts, or specific split files).
Hardware Specification Yes We performed our experimental evaluation on a machine equipped with an NVIDA RTX 2080ti, within the Pytorch framework.
Software Dependencies No The paper mentions using the "Pytorch framework" but does not specify its version number or any other software dependencies with their versions.
Experiment Setup Yes in all our experiments we use a latent space of dimension d = 10, unless otherwise specified, and k = 10 latent subspaces.