Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]

Convergence of denoising diffusion models under the manifold hypothesis

Authors: Valentin De Bortoli

TMLR 2022 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Despite their strong empirical results, the theoretical analysis of such models remains limited. In particular, all current approaches crucially assume that the target density admits a density w.r.t. the Lebesgue measure. [...] In this paper, we bridge this gap by providing the first convergence results for diffusion models in this setting. In particular, we provide quantitative bounds on the Wasserstein distance of order one between the target data distribution and the generative distribution of the diffusion model.
Researcher Affiliation Academia Valentin De Bortoli Department of Computer Science ENS, CNRS, PSL University Paris, France
Pseudocode No The paper describes mathematical equations and processes, such as equation (5) for the EI discretization, and outlines steps in prose. However, it does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks, nor does it present structured, code-like procedural steps.
Open Source Code No The paper states: 'Reviewed on Open Review: https: // openreview .net/ forum? id= Mh K5a Xo3g B&', which is a link to the open review platform, not a repository for source code. There is no explicit statement or link in the paper indicating the availability of source code for the described methodology.
Open Datasets No The paper discusses theoretical concepts related to data distributions, such as the manifold hypothesis and empirical measures (e.g., 'empirical measures of the form (1/N) PN i=1 δXi'). It mentions applications in 'image and audio synthesis' and 'image processing' as motivation for the theoretical work. However, the paper is theoretical and does not conduct experiments with specific datasets, therefore, it does not provide access information for any publicly available or open datasets used in empirical studies.
Dataset Splits No The paper is theoretical and does not present any empirical experiments involving dataset usage. Therefore, there is no mention or description of training/test/validation dataset splits.
Hardware Specification No The paper is a theoretical work focusing on convergence results for diffusion models. It does not describe any experiments that would require specific hardware. Consequently, there is no mention of GPU models, CPU types, or other hardware specifications used for running experiments.
Software Dependencies No The paper is a theoretical analysis of denoising diffusion models and does not involve practical implementation or empirical experiments. Therefore, it does not list any specific software components or their version numbers that would be required to reproduce experimental results.
Experiment Setup No The paper is purely theoretical, focusing on mathematical proofs and convergence bounds for diffusion models. It does not describe any empirical experiments, and therefore, no experimental setup details, hyperparameters, or training configurations are provided in the text.