解构去噪扩散模型用于自监督学习

Deconstructing Denoising Diffusion Models for Self-Supervised Learning

January 25, 2024

作者: Xinlei Chen, Zhuang Liu, Saining Xie, Kaiming He

cs.AI

摘要

在这项研究中，我们研究了最初用于图像生成的去噪扩散模型（Denoising Diffusion Models，DDM）的表示学习能力。我们的理念是对DDM进行解构，逐渐将其转变为经典的去噪自编码器（Denoising Autoencoder，DAE）。这种解构过程使我们能够探索现代DDM的各种组件如何影响自监督表示学习。我们观察到，只有很少一部分现代组件对于学习良好的表示是至关重要的，而许多其他组件则是非必要的。我们的研究最终得出了一种高度简化的方法，在很大程度上类似于经典的DAE。我们希望我们的研究能重新激起人们对现代自监督学习领域内一类经典方法的兴趣。

English

In this study, we examine the representation learning abilities of Denoising Diffusion Models (DDM) that were originally purposed for image generation. Our philosophy is to deconstruct a DDM, gradually transforming it into a classical Denoising Autoencoder (DAE). This deconstructive procedure allows us to explore how various components of modern DDMs influence self-supervised representation learning. We observe that only a very few modern components are critical for learning good representations, while many others are nonessential. Our study ultimately arrives at an approach that is highly simplified and to a large extent resembles a classical DAE. We hope our study will rekindle interest in a family of classical methods within the realm of modern self-supervised learning.