Learning Mixtures of Gaussians Using the DDPM Objective

Authors: Kulin Shah, Sitan Chen, Adam Klivans

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work, we give the first provably efficient results along these lines for one of the most fundamental distribution families, Gaussian mixture models. We prove that gradient descent on the denoising diffusion probabilistic model (DDPM) objective can efficiently recover the ground truth parameters of the mixture model in the following two settings: 1. We show gradient descent with random initialization learns mixtures of two spherical Gaussians in d dimensions with 1/poly(d)-separated centers. 2. We show gradient descent with a warm start learns mixtures of K spherical Gaussians with (log(min(K, d)))-separated centers.
Researcher Affiliation Academia UT Austin kulinshah@utexas.edu Sitan Chen Harvard University sitan@seas.harvard.edu Adam Klivans UT Austin klivans@cs.utexas.edu
Pseudocode Yes Algorithm 1: GMMDENOISER(t, {µ(0)i}K i=1, H)
Open Source Code No The paper is theoretical and does not mention any open-source code for the described methodology.
Open Datasets No The paper is theoretical and focuses on proofs and algorithms. It refers to data distributions and samples in a theoretical context but does not mention the use of any specific publicly available datasets for training or provide access information for such datasets.
Dataset Splits No The paper is theoretical and does not discuss dataset splits for training, validation, or testing.
Hardware Specification No The paper is theoretical and does not describe any specific hardware used for running experiments.
Software Dependencies No The paper is theoretical and does not mention any specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and focuses on mathematical proofs and algorithmic analysis. It does not describe an experimental setup, hyperparameters, or training configurations.