The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models

• URL: http://arxiv.org/abs/2605.06367v1
• Date: Thu, 07 May 2026 14:44:53 GMT
• Title: The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models
• Abstract summary: We study class-dependent learning in score-based diffusion models.<n>We reveal the explicit hierarchy governing these dynamics.<n>We validate our theoretical predictions empirically using U-Net models trained on Fashion MNIST.
• Score: 8.547042933482645
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Real-world datasets are inherently heterogeneous, yet how per-class structural differences and sampling imbalance shape the training dynamics of diffusion models-and potentially exacerbate disparities-remains poorly understood. While models typically transition from an initial phase of generalization to memorizing the training set, existing theory assumes homogeneous data, leaving open how class imbalance and heterogeneity reshape these dynamics. In this work, we develop a high-dimensional analytical framework to study class-dependent learning in score-based diffusion models. Analyzing a random-features model trained on Gaussian mixtures, we derive the feature-covariance spectrum to characterize per-class generalization and memorization times. We reveal the explicit hierarchy governing these dynamics: class variance is the primary determinant of learning order-consistently favoring higher-variance classes-while centroid geometry plays a secondary role. Sampling imbalance acts as a modulator that can reverse this ordering and, under strong imbalance, forces minority classes to acquire distinct, delayed speciation times during backward diffusion. Together, these results suggest that diffusion models can memorize some classes while others remain insufficiently learned. We validate our theoretical predictions empirically using U-Net models trained on Fashion MNIST.

Related papers

• Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refine [60.669081685261965]
  Diffusion models generate high-dimensional data with remarkable quality.<n>How their training efficiently learns the score function remains theoretically unexplained.<n>We instantiate this principle as Score-induced Latent Diffusion (SiLD)
  arXiv  Detail & Related papers  (2026-05-16T16:51:10Z)
• Diffusion Model for Manifold Data: Score Decomposition, Curvature, and Statistical Complexity [93.52679474816728]
  Diffusion models have become a leading framework in generative modeling.<n>This paper investigates how diffusion models learn structured data, focusing on two key aspects: statistical complexity and influence of data geometric properties.<n>We also highlight the interplay of manifold curvature with the structures in the score function.
  arXiv  Detail & Related papers  (2026-03-21T04:40:19Z)
• How Out-of-Equilibrium Phase Transitions can Seed Pattern Formation in Trained Diffusion Models [12.52425103385255]
  We argue that reverse diffusion passes through a critical regime in which small spatial fluctuations are amplified and seed the emergence of large-scale structure.<n>Using analytically tractable patch score models, we show how classical symmetry-breaking bifurcations generalize into spatially extended critical phenomena.
  arXiv  Detail & Related papers  (2026-03-20T16:19:10Z)
• Theory of Speciation Transitions in Diffusion Models with General Class Structure [5.939780039158003]
  Diffusion models generate data by reversing a diffusion process, transforming noise into structured samples drawn from a target distribution.<n>Recent theoretical work has shown that this backward dynamics can undergo sharp qualitative transitions, known as speciation transitions.<n>We develop a general theory of speciation in diffusion models that applies to arbitrary target distributions admitting well-defined classes.
  arXiv  Detail & Related papers  (2026-02-04T10:35:47Z)
• On the Edge of Memorization in Diffusion Models [25.927892368310868]
  We introduce a scientific and mathematical "laboratory" for investigating memorization and generalization in practical diffusion models.<n>Our work provides an analytically tractable and practically meaningful setting for future theoretical and empirical investigations.
  arXiv  Detail & Related papers  (2025-08-25T05:56:05Z)
• Overcoming Dimensional Factorization Limits in Discrete Diffusion Models through Quantum Joint Distribution Learning [79.65014491424151]
  We propose a quantum Discrete Denoising Diffusion Probabilistic Model (QD3PM)<n>It enables joint probability learning through diffusion and denoising in exponentially large Hilbert spaces.<n>This paper establishes a new theoretical paradigm in generative models by leveraging the quantum advantage in joint distribution learning.
  arXiv  Detail & Related papers  (2025-05-08T11:48:21Z)
• On the Feature Learning in Diffusion Models [26.53807235141923]
  We propose a feature learning framework aimed at analyzing and comparing the training dynamics of diffusion models with those of traditional classification models.<n>Our theoretical analysis demonstrates that diffusion models, due to the denoising objective, are encouraged to learn more balanced and comprehensive representations of the data.<n>In contrast, neural networks with a similar architecture trained for classification tend to prioritize learning specific patterns in the data, often focusing on easy-to-learn components.
  arXiv  Detail & Related papers  (2024-12-02T00:41:25Z)
• Bias in Motion: Theoretical Insights into the Dynamics of Bias in SGD Training [7.5041863920639456]
  Machine learning systems often acquire biases by leveraging undesired features in the data, impacting accuracy across different sub-populations.<n>This paper explores the evolution of bias in a teacher-student setup modeling different data sub-populations with a Gaussian-mixture model.<n>Applying our findings to fairness and robustness, we delineate how and when heterogeneous data and spurious features can generate and amplify bias.
  arXiv  Detail & Related papers  (2024-05-28T15:50:10Z)
• DIFFormer: Scalable (Graph) Transformers Induced by Energy Constrained Diffusion [66.21290235237808]
  We introduce an energy constrained diffusion model which encodes a batch of instances from a dataset into evolutionary states. We provide rigorous theory that implies closed-form optimal estimates for the pairwise diffusion strength among arbitrary instance pairs. Experiments highlight the wide applicability of our model as a general-purpose encoder backbone with superior performance in various tasks.
  arXiv  Detail & Related papers  (2023-01-23T15:18:54Z)
• On the Generalization and Adaption Performance of Causal Models [99.64022680811281]
  Differentiable causal discovery has proposed to factorize the data generating process into a set of modules. We study the generalization and adaption performance of such modular neural causal models. Our analysis shows that the modular neural causal models outperform other models on both zero and few-shot adaptation in low data regimes.
  arXiv  Detail & Related papers  (2022-06-09T17:12:32Z)
• Why do classifier accuracies show linear trends under distribution shift? [58.40438263312526]
  accuracies of models on one data distribution are approximately linear functions of the accuracies on another distribution. We assume the probability that two models agree in their predictions is higher than what we can infer from their accuracy levels alone. We show that a linear trend must occur when evaluating models on two distributions unless the size of the distribution shift is large.
  arXiv  Detail & Related papers  (2020-12-31T07:24:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.