Evaluating the design space of diffusion-based generative models

Authors: Yuqing Wang, Ye He, Molei Tao

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical This article instead provides a first quantitative understanding of the whole generation process, i.e., both training and sampling. More precisely, it conducts a non-asymptotic convergence analysis of denoising score matching under gradient descent. In addition, a refined sampling error analysis for variance exploding models is also provided. The combination of these two results yields a full error analysis... and 'No experiments.' in the checklist.
Researcher Affiliation Academia Yuqing Wang Simons Institute University of California, Berkeley yq.wang@berkeley.edu Ye He School of Mathematics Georgia Institute of Technology yhe367@gatech.edu Molei Tao School of Mathematics Georgia Institute of Technology mtao@gatech.edu
Pseudocode No The paper describes mathematical and algorithmic processes, such as 'exponential integrator scheme for simulating the backward SDE', but does not contain a clearly labeled 'Pseudocode' or 'Algorithm' block.
Open Source Code No The 'Experimental Result Reproducibility' and 'Open access to data and code' sections in the NeurIPS checklist are marked 'NA' and explicitly state 'No experiments.', indicating no code release.
Open Datasets No The NeurIPS checklist section 'Experimental Result Reproducibility' explicitly states 'No experiments.', indicating no datasets were used.
Dataset Splits No The NeurIPS checklist explicitly states 'No experiments.', thus no dataset splits are provided.
Hardware Specification No The NeurIPS checklist section 'Experiments Compute Resources' is marked 'NA' and states 'No experiments.', indicating no specific hardware details are provided.
Software Dependencies No The NeurIPS checklist explicitly states 'No experiments.', thus no software dependencies with version numbers are listed.
Experiment Setup No The NeurIPS checklist section 'Experimental Setting/Details' is marked 'NA' and states 'No experiments.', indicating no specific experimental setup details are provided.