Understanding Classifier-Free Guidance: High-Dimensional Theory and Non-Linear Generalizations

• URL: http://arxiv.org/abs/2502.07849v1
• Date: Tue, 11 Feb 2025 10:29:29 GMT
• Title: Understanding Classifier-Free Guidance: High-Dimensional Theory and Non-Linear Generalizations
• Authors: Krunoslav Lehman Pavasovic, Jakob Verbeek, Giulio Biroli, Marc Mezard,
• Abstract summary: We show thatconditional-Free Guidance effectively reproduces the target distribution in infinite and sufficiently high-dimensional contexts. We also explore finite-dimensional effects, precisely characterizing overshoot and variance reduction.
• Score: 22.44946627454133
• License:
• Abstract: Recent studies have raised concerns about the effectiveness of Classifier-Free Guidance (CFG), indicating that in low-dimensional settings, it can lead to overshooting the target distribution and reducing sample diversity. In this work, we demonstrate that in infinite and sufficiently high-dimensional contexts CFG effectively reproduces the target distribution, revealing a blessing-of-dimensionality result. Additionally, we explore finite-dimensional effects, precisely characterizing overshoot and variance reduction. Based on our analysis, we introduce non-linear generalizations of CFG. Through numerical simulations on Gaussian mixtures and experiments on class-conditional and text-to-image diffusion models, we validate our analysis and show that our non-linear CFG offers improved flexibility and generation quality without additional computation cost.

Related papers

• Nested Annealed Training Scheme for Generative Adversarial Networks [54.70743279423088]
  This paper focuses on a rigorous mathematical theoretical framework: the composite-functional-gradient GAN (CFG) We reveal the theoretical connection between the CFG model and score-based models. We find that the training objective of the CFG discriminator is equivalent to finding an optimal D(x)
  arXiv  Detail & Related papers  (2025-01-20T07:44:09Z)
• A New Formulation of Lipschitz Constrained With Functional Gradient Learning for GANs [52.55025869932486]
  This paper introduces a promising alternative method for training Generative Adversarial Networks (GANs) on large-scale datasets with clear theoretical guarantees. We propose a novel Lipschitz-constrained Functional Gradient GANs learning (Li-CFG) method to stabilize the training of GAN. We demonstrate that the neighborhood size of the latent vector can be reduced by increasing the norm of the discriminator gradient.
  arXiv  Detail & Related papers  (2025-01-20T02:48:07Z)
• Contrastive CFG: Improving CFG in Diffusion Models by Contrasting Positive and Negative Concepts [55.298031232672734]
  As-Free Guidance (CFG) has proven effective in conditional diffusion model sampling for improved condition alignment. We present a novel method to enhance negative CFG guidance using contrastive loss.
  arXiv  Detail & Related papers  (2024-11-26T03:29:27Z)
• A PAC-Bayesian Perspective on the Interpolating Information Criterion [54.548058449535155]
  We show how a PAC-Bayes bound is obtained for a general class of models, characterizing factors which influence performance in the interpolating regime. We quantify how the test error for overparameterized models achieving effectively zero training error depends on the quality of the implicit regularization imposed by e.g. the combination of model, parameter-initialization scheme.
  arXiv  Detail & Related papers  (2023-11-13T01:48:08Z)
• Boosted Control Functions: Distribution generalization and invariance in confounded models [10.503777692702952]
  We introduce a strong notion of invariance that allows for distribution generalization even in the presence of nonlinear, non-identifiable structural functions. We propose the ControlTwicing algorithm to estimate the Boosted Control Function (BCF) using flexible machine-learning techniques.
  arXiv  Detail & Related papers  (2023-10-09T15:43:46Z)
• On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data [26.310275682709776]
  Various logit-adjusted parameterizations of the cross-entropy (CE) loss have been proposed as alternatives to weighted CE large models on labelimbalanced data. We show that logit-adjusted parameterizations can be appropriately tuned to learn to learn irrespective of the minority imbalance ratio.
  arXiv  Detail & Related papers  (2023-03-14T03:04:37Z)
• Causal Graph Discovery from Self and Mutually Exciting Time Series [10.410454851418548]
  We develop a non-asymptotic recovery guarantee and quantifiable uncertainty by solving a linear program. We demonstrate the effectiveness of our approach in recovering highly interpretable causal DAGs over Sepsis Associated Derangements (SADs)
  arXiv  Detail & Related papers  (2023-01-26T16:15:27Z)
• Pseudo-Spherical Contrastive Divergence [119.28384561517292]
  We propose pseudo-spherical contrastive divergence (PS-CD) to generalize maximum learning likelihood of energy-based models. PS-CD avoids the intractable partition function and provides a generalized family of learning objectives.
  arXiv  Detail & Related papers  (2021-11-01T09:17:15Z)
• Causal Graph Discovery from Self and Mutually Exciting Time Series [12.802653884445132]
  We develop a non-asymptotic recovery guarantee and quantifiable uncertainty by solving a linear program. We demonstrate the effectiveness of our approach in recovering highly interpretable causal DAGs over Sepsis Associated Derangements (SADs)
  arXiv  Detail & Related papers  (2021-06-04T16:59:24Z)
• On the Benefits of Invariance in Neural Networks [56.362579457990094]
  We show that training with data augmentation leads to better estimates of risk and thereof gradients, and we provide a PAC-Bayes generalization bound for models trained with data augmentation. We also show that compared to data augmentation, feature averaging reduces generalization error when used with convex losses, and tightens PAC-Bayes bounds.
  arXiv  Detail & Related papers  (2020-05-01T02:08:58Z)

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.