A Good Score Does not Lead to A Good Generative Model

• URL: http://arxiv.org/abs/2401.04856v2
• Date: Sat, 27 Jan 2024 17:42:19 GMT
• Title: A Good Score Does not Lead to A Good Generative Model
• Authors: Sixu Li, Shi Chen, Qin Li
• Abstract summary: Score-based Generative Models (SGMs) is one leading method in generative modeling. We show that SGMs can generate samples from a distribution that is close to the ground-truth if the underlying score function is learned well.
• Score: 14.752242187781107
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Score-based Generative Models (SGMs) is one leading method in generative modeling, renowned for their ability to generate high-quality samples from complex, high-dimensional data distributions. The method enjoys empirical success and is supported by rigorous theoretical convergence properties. In particular, it has been shown that SGMs can generate samples from a distribution that is close to the ground-truth if the underlying score function is learned well, suggesting the success of SGM as a generative model. We provide a counter-example in this paper. Through the sample complexity argument, we provide one specific setting where the score function is learned well. Yet, SGMs in this setting can only output samples that are Gaussian blurring of training data points, mimicking the effects of kernel density estimation. The finding resonates a series of recent finding that reveal that SGMs can demonstrate strong memorization effect and fail to generate.

Related papers

• Wasserstein Convergence Guarantees for a General Class of Score-Based Generative Models [9.47767039367222]
  Score-based generative models (SGMs) are a recent class of deep generative models with state-of-the-art performance in many applications. We establish convergence guarantees for a general class of SGMs in 2-Wasserstein distance, assuming accurate score estimates and smooth log-concave data distribution. Numerically, we experiment SGMs with different forward processes, some of which are newly proposed in this paper, for unconditional image generation on CIFAR-10.
  arXiv  Detail & Related papers  (2023-11-18T07:53:22Z)
• Self-Supervised Dataset Distillation for Transfer Learning [77.4714995131992]
  We propose a novel problem of distilling an unlabeled dataset into a set of small synthetic samples for efficient self-supervised learning (SSL) We first prove that a gradient of synthetic samples with respect to a SSL objective in naive bilevel optimization is textitbiased due to randomness originating from data augmentations or masking. We empirically validate the effectiveness of our method on various applications involving transfer learning.
  arXiv  Detail & Related papers  (2023-10-10T10:48:52Z)
• Accurate generation of stochastic dynamics based on multi-model Generative Adversarial Networks [0.0]
  Generative Adversarial Networks (GANs) have shown immense potential in fields such as text and image generation. Here we quantitatively test this approach by applying it to a prototypical process on a lattice. Importantly, the discreteness of the model is retained despite the noise.
  arXiv  Detail & Related papers  (2023-05-25T10:41:02Z)
• Convergence for score-based generative modeling with polynomial complexity [9.953088581242845]
  We prove the first convergence guarantees for the core mechanic behind Score-based generative modeling. Compared to previous works, we do not incur error that grows exponentially in time or that suffers from a curse of dimensionality. We show that a predictor-corrector gives better convergence than using either portion alone.
  arXiv  Detail & Related papers  (2022-06-13T14:57:35Z)
• Score-Based Generative Models Detect Manifolds [0.0]
  Score-based generative models (SGMs) need to approximate the scores $nabla log p_t$ of the intermediate distributions as well as the final distribution $p_T$ of the forward process. We find precise conditions under which SGMs are able to produce samples from an underlying (low-dimensional) data manifold.
  arXiv  Detail & Related papers  (2022-06-02T12:29:10Z)
• Sampling from Arbitrary Functions via PSD Models [55.41644538483948]
  We take a two-step approach by first modeling the probability distribution and then sampling from that model. We show that these models can approximate a large class of densities concisely using few evaluations, and present a simple algorithm to effectively sample from these models.
  arXiv  Detail & Related papers  (2021-10-20T12:25:22Z)
• Score-based Generative Modeling in Latent Space [93.8985523558869]
  Score-based generative models (SGMs) have recently demonstrated impressive results in terms of both sample quality and distribution coverage. Here, we propose the Latent Score-based Generative Model (LSGM), a novel approach that trains SGMs in a latent space. Moving from data to latent space allows us to train more expressive generative models, apply SGMs to non-continuous data, and learn smoother SGMs in a smaller space.
  arXiv  Detail & Related papers  (2021-06-10T17:26:35Z)
• Continual Learning with Fully Probabilistic Models [70.3497683558609]
  We present an approach for continual learning based on fully probabilistic (or generative) models of machine learning. We propose a pseudo-rehearsal approach using a Gaussian Mixture Model (GMM) instance for both generator and classifier functionalities. We show that GMR achieves state-of-the-art performance on common class-incremental learning problems at very competitive time and memory complexity.
  arXiv  Detail & Related papers  (2021-04-19T12:26:26Z)
• Robust Finite Mixture Regression for Heterogeneous Targets [70.19798470463378]
  We propose an FMR model that finds sample clusters and jointly models multiple incomplete mixed-type targets simultaneously. We provide non-asymptotic oracle performance bounds for our model under a high-dimensional learning framework. The results show that our model can achieve state-of-the-art performance.
  arXiv  Detail & Related papers  (2020-10-12T03:27:07Z)
• Efficiently Sampling Functions from Gaussian Process Posteriors [76.94808614373609]
  We propose an easy-to-use and general-purpose approach for fast posterior sampling. We demonstrate how decoupled sample paths accurately represent Gaussian process posteriors at a fraction of the usual cost.
  arXiv  Detail & Related papers  (2020-02-21T14:03:16Z)

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.