Disentangling by Subspace Diffusion

Authors: David Pfau, Irina Higgins, Alex Botev, Sébastien Racanière

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To demonstrate the power of GEOMANCER, we investigate its performance on both synthetic manifolds and a dataset of rendered 3D objects.
Researcher Affiliation Industry David Pfau, Irina Higgins, Aleksandar Botev, Sébastien Racanière Deep Mind London, UK {pfau, irinah, botev, sracaniere}@google.com
Pseudocode Yes Algorithm 1: Geometric Manifold Component Estimator (GEOMANCER)
Open Source Code No The paper does not provide any explicit statements about the release of its source code or links to a code repository.
Open Datasets Yes Stanford 3D Scanning Repository [71]
Dataset Splits No The paper describes how data was generated or sampled for experiments (e.g., 'uniformly sampling from either the n-dimensional sphere Sn Rn+1', 'sampled rotations of object pose uniformly from SO(3)'), but it does not specify explicit training, validation, or test dataset splits (e.g., percentages, counts, or references to predefined splits) for the data used to evaluate GEOMANCER.
Hardware Specification No The paper does not specify any hardware details (e.g., CPU, GPU models, memory, or cloud resources) used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks like 'Python 3.8' or 'PyTorch 1.9').
Experiment Setup Yes GEOMANCER requires very few hyperparameters just the dimension k, the number of nearest neighbors, and the gap γ in the spectrum of 2 at which to stop splitting tangent spaces, which can be chosen by simple heuristics.