Recent advances in deep generative models have led to impressive results in a variety of application domains. Motivated by the possibility that deep learning models might memorize part of the input data, there have been increased efforts to understand how memorization arises. In this work, we extend a recently proposed measure of memorization for supervised learning (Feldman, 2019) to the unsupervised density estimation problem and adapt it to be more computationally efficient. Next, we present a study that demonstrates how memorization can occur in probabilistic deep generative models such as variational autoencoders. This reveals that the form of memorization to which these models are susceptible differs fundamentally from mode collapse and overfitting. Furthermore, we show that the proposed memorization score measures a phenomenon that is not captured by commonly-used nearest neighbor tests. Finally, we discuss several strategies that can be used to limit memorization in practice. Our work thus provides a framework for understanding problematic memorization in probabilistic generative models.
Losing dimensions: Geometric memorization in generative diffusion
Computing Sciences Department, Bocconi University（博科尼大学计算科学系）；Bocconi Institute for Data Science and Analytics (BIDSA)（博科尼数据科学与分析研究所）；Donders Institute, Radboud University（拉德堡德大学多纳尔斯研究所）；OnePlanet Research Center, Wageningen（瓦赫宁根大学OnePlanet研究中心）；Rensselaer Polytechnic Institute（伦斯勒理工学院）；Jheronimus Academy of Data Science (JADS)（杰霍尼穆斯数据科学学院）；IBM Research, Cambridge（IBM研究院，剑桥）
Models for learning probability distributions such as generative models and density estimators behave quite differently from models for learning functions. One example is found in the memorization phenomenon, namely the ultimate convergence to the empirical distribution, that occurs in generative adversarial networks (GANs). For this reason, the issue of generalization is more subtle than that for supervised learning. For the bias potential model, we show that dimension-independent generalization accuracy is achievable if early stopping is adopted, despite that in the long term, the model eith…
Understanding and Mitigating Memorization in Generative Models via Sharpness of Probability Landscapes
Department of Computer Science, Yonsei University, Seoul, Korea（计算机科学系，延世大学，首尔，韩国）；Department of Artificial Intelligence, Yonsei University, Seoul, Korea（人工智能系，延世大学，首尔，韩国）
University of California, Los Angeles (UCLA)（加州大学洛杉矶分校）；Google（谷歌）；National Taiwan University（台湾大学）
Generative adversarial network (GAN) continues to be a popular research direction due to its high generation quality. It is observed that many state-of-the-art GANs generate samples that are more similar to the training set than a holdout testing set from the same distribution, hinting some training samples are implicitly memorized in these models. This memorization behavior is unfavorable in many applications that demand the generated samples to be sufficiently distinct from known samples. Nevertheless, it is unclear whether it is possible to reduce memorization without compromising the gener…
Normandie Univ（诺曼底大学）；UNICAEN（卡昂大学）；ENSICAEN（卡昂高等工程师学院）；CNRS — UMR GREYC（法国国家科学研究中心 — 灰谐实验室联合研究单位（注：CNRS为法国国家科学研究中心，UMR为联合研究单位缩写，GREYC为实验室名，此处按原文提取机构部分））
State of the art deep generative networks are capable of producing images with such incredible realism that they can be suspected of memorizing training images. It is why it is not uncommon to include visualizations of training set nearest neighbors, to suggest generated images are not simply memorized. We demonstrate this is not sufficient and motivates the need to study memorization/overfitting of deep generators with more scrutiny. This paper addresses this question by i) showing how simple losses are highly effective at reconstructing images for deep generators ii) analyzing the statistics…
On the Memorization of Consistency Distillation for Diffusion Models
School of Computing & Data Science, The University of Hong Kong（计算与数据科学学院，香港大学）；School of Computing & Data Science and Institute of Data Science, The University of Hong Kong（计算与数据科学学院和数据科学研究所，香港大学）
We propose an approach to address two issues that commonly occur during training of unsupervised GANs. First, since GANs use only a continuous latent distribution to embed multiple classes or clusters of data, they often do not correctly handle the structural discontinuity between disparate classes in a latent space. Second, discriminators of GANs easily forget about past generated samples by generators, incurring instability during adversarial training. We argue that these two infamous problems of unsupervised GAN training can be largely alleviated by a learnable memory network to which both …
Reducing Diffusion Model Memorization with Higher Order Langevin Dynamics
Department of Applied Math & Statistics（应用数学与统计学系）；Stony Brook University（石溪大学）；Department of Electrical and Computer Engineering（电气与计算机工程系）；Faculty of Engineering（工程学院）；Bar-Ilan University（巴伊兰大学）
Memorization to Generalization: Emergence of Diffusion Models from Associative Memory
Department of Computer Science, Rensselaer Polytechnic Institute ( RPI )（计算机科学系，罗切斯特理工学院（RPI））；Jheronimus Academy of Data Science, Tilburg University（数据科学学院，蒂尔堡大学）；Department of Physics, University of Rome Sapienza（物理系，罗马萨皮恩扎大学）；Donders Institute for Brain, Cognition, and Behaviour, Radboud University（大脑、认知与行为研究所，拉德堡德大学）；IBM Research（IBM研究院）
Loss Curvature as a Sampling-Free Proxy for Memorization in Diffusion Models
Department of Mathematics, University of California, Los Angeles（加州大学洛杉矶分校数学系）；Elmore Family School of Electrical and Computer Engineering, Purdue University（普渡大学埃尔默电气与计算机工程学院）
We introduce the Probabilistic Generative Adversarial Network (PGAN), a new GAN variant based on a new kind of objective function. The central idea is to integrate a probabilistic model (a Gaussian Mixture Model, in our case) into the GAN framework which supports a new kind of loss function (based on likelihood rather than classification loss), and at the same time gives a meaningful measure of the quality of the outputs generated by the network. Experiments with MNIST show that the model learns to generate realistic images, and at the same time computes likelihoods that are correlated with th…
Unconditional Latent Diffusion Models Memorize Patient Imaging Data: Implications for Openly Sharing Synthetic Data
Heidelberg University Hospital（海德堡大学医院）；AI Health Innovation Cluster (AIH)（AI健康创新集群）；German Centre for Cardiovascular Research (DZHK)（德国心血管研究中心）；University Medical Center Mannheim（曼海姆大学医学中心）；University Hospital Aachen（亚琛大学医院）；University Hospital Würzburg（维尔茨堡大学医院）；University Medical Center Mainz（美因茨大学医学中心）；TUD Dresden University of Technology（德累斯顿工业大学）；University Hospital Dresden（德累斯顿大学医院）；National Center for Tumor Diseases (NCT)（国家肿瘤疾病中心）
Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neur…
Federal University of Rio de Janeiro（里约热内卢联邦大学）；Queen Mary University of London（伦敦玛丽女王大学）；Barcelona Supercomputing Center（巴塞罗那超级计算中心）
Accurate determination of fuel properties of complex mixtures over a wide range of pressure and temperature conditions is essential to utilizing alternative fuels. The present work aims to construct cheap-to-compute machine learning (ML) models to act as closure equations for predicting the physical properties of alternative fuels. Those models can be trained using the database from MD simulations and/or experimental measurements in a data-fusion-fidelity approach. Here, Gaussian Process (GP) and probabilistic generative models are adopted. GP is a popular non-parametric Bayesian approach to b…
University of Kaiserslautern（凯泽斯劳滕大学）；German Research Center for Artificial Intelligence (DFKI)（德国人工智能研究中心（DFKI））
Time series forecasting is one of the challenging problems for humankind. Traditional forecasting methods using mean regression models have severe shortcomings in reflecting real-world fluctuations. While new probabilistic methods rush to rescue, they fight with technical difficulties like quantile crossing or selecting a prior distribution. To meld the different strengths of these fields while avoiding their weaknesses as well as to push the boundary of the state-of-the-art, we introduce ForGAN - one step ahead probabilistic forecasting with generative adversarial networks. ForGAN utilizes th…
It can be difficult to tell whether a trained generative model has learned to generate novel examples or has simply memorized a specific set of outputs. In published work, it is common to attempt to address this visually, for example by displaying a generated example and its nearest neighbor(s) in the training set (in, for example, the L2 metric). As any generative model induces a probability density on its output domain, we propose studying this density directly. We first study the geometry of the latent representation and generator, relate this to the output density, and then develop techniq…
Chungbuk National University Hospital（忠北国立大学医院）；Korea University（高丽大学）；Hannam University（韩南大学）
This paper presents a novel convolutional layer, called perturbed convolution (PConv), which focuses on achieving two goals simultaneously: improving the generative adversarial network (GAN) performance and alleviating the memorization problem in which the discriminator memorizes all images from a given dataset as training progresses. In PConv, perturbed features are generated by randomly disturbing an input tensor before performing the convolution operation. This approach is simple but surprisingly effective. First, to produce a similar output even with the perturbed tensor, each layer in the…
Deep neural networks have made breakthroughs in a wide range of visual understanding tasks. A typical challenge that hinders their real-world applications is that unknown samples may be fed into the system during the testing phase, but traditional deep neural networks will wrongly recognize these unknown samples as one of the known classes. Open set recognition (OSR) is a potential solution to overcome this problem, where the open set classifier should have the flexibility to reject unknown samples and meanwhile maintain high classification accuracy in known classes. Probabilistic generative m…
A Probabilistic Fluctuation based Membership Inference Attack for Diffusion Models
School of Cyber Science and Engineering, Huazhong University of Science and Technology（华中科技大学网络空间科学与工程学院）；Tsinghua University（清华大学）；Dalian University of Technology（大连理工大学）
National Taiwan University（台湾大学）；Google（谷歌）；Kaggle（卡格尔）
Many recent developments on generative models for natural images have relied on heuristically-motivated metrics that can be easily gamed by memorizing a small sample from the true distribution or training a model directly to improve the metric. In this work, we critically evaluate the gameability of these metrics by designing and deploying a generative modeling competition. Our competition received over 11000 submitted models. The competitiveness between participants allowed us to investigate both intentional and unintentional memorization in generative modeling. To detect intentional memoriza…
Real-time probabilistic tsunami forecasting via generative AI
Fujitsu Research, Fujitsu Limited（富士通研究所，富士通株式会社）；Earthquake Research Institute, The University of Tokyo（东京大学地震研究所）；International Research Institute of Disaster Science, Tohoku University（东北大学国际灾害科学研究所）
MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）；Google Research（谷歌研究院）
We study the problem of 3D object generation. We propose a novel framework, namely 3D Generative Adversarial Network (3D-GAN), which generates 3D objects from a probabilistic space by leveraging recent advances in volumetric convolutional networks and generative adversarial nets. The benefits of our model are three-fold: first, the use of an adversarial criterion, instead of traditional heuristic criteria, enables the generator to capture object structure implicitly and to synthesize high-quality 3D objects; second, the generator establishes a mapping from a low-dimensional probabilistic space…
NORCE Norwegian Research Centre（NORCE挪威研究中心）；Heriot-Watt University（赫瑞-瓦特大学）
Quantitative workflows utilizing real-time data to constrain ahead-of-bit uncertainty have the potential to improve geosteering significantly. Fast updates based on real-time data are essential when drilling in complex reservoirs with high uncertainties in pre-drill models. However, practical assimilation of real-time data requires effective geological modeling and mathematically robust parameterization. We propose a generative adversarial deep neural network (GAN), trained to reproduce geologically consistent 2D sections of fluvial successions. Offline training produces a fast GAN-based appro…
Greater direct electrification of end-use sectors with a higher share of renewables is one of the pillars to power a carbon-neutral society by 2050. However, in contrast to conventional power plants, renewable energy is subject to uncertainty raising challenges for their interaction with power systems. Scenario-based probabilistic forecasting models have become a vital tool to equip decision-makers. This paper presents to the power systems forecasting practitioners a recent deep learning technique, the normalizing flows, to produce accurate scenario-based probabilistic forecasts that are cruci…
Tsinghua University（清华大学）；University of California, Los Angeles（加州大学洛杉矶分校）；Peking University（北京大学）；Beijing Institute for General Artificial Intelligence（北京通用人工智能研究院）
Probabilistic Circuits (PCs) are a general and unified computational framework for tractable probabilistic models that support efficient computation of various inference tasks (e.g., computing marginal probabilities). Towards enabling such reasoning capabilities in complex real-world tasks, Liu et al. (2022) propose to distill knowledge (through latent variable assignments) from less tractable but more expressive deep generative models. However, it is still unclear what factors make this distillation work well. In this paper, we theoretically and empirically discover that the performance of a …
Efficient Generative AI Boosts Probabilistic Forecasting of Sudden Stratospheric Warmings
Beijing Normal University（北京师范大学）；Institute of Atmospheric Physics, Chinese Academy of Sciences（中国科学院大气物理研究所）；Zhejiang University（浙江大学）；Chinese Academy of Meteorological Sciences（中国气象科学研究院）；Lanzhou University（兰州大学）
Carnegie Mellon University（卡内基梅隆大学）；DeepMind（深度思维）；University of California, San Diego（加利福尼亚大学圣迭戈分校）
We propose a deep and interpretable probabilistic generative model to analyze glyph shapes in printed Early Modern documents. We focus on clustering extracted glyph images into underlying templates in the presence of multiple confounding sources of variance. Our approach introduces a neural editor model that first generates well-understood printing phenomena like spatial perturbations from template parameters via interpertable latent variables, and then modifies the result by generating a non-interpretable latent vector responsible for inking variations, jitter, noise from the archiving proces…
Wasserstein proximal operators describe score-based generative models and resolve memorization
University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）；University of South Carolina, Columbia（南卡罗来纳大学哥伦比亚分校）
A survey of probabilistic generative frameworks for molecular simulations
University of Maryland（马里兰大学）；Institute for Physical Science and Technology（物理科学与技术研究所）；University of Maryland Institute for Health Computing（马里兰大学健康计算研究所）；Department of Chemistry and Biochemistry（化学与生物化学系）
IBM T.J. Watson Research Center（IBM托马斯·J·沃森研究中心）；Rensselaer Polytechnic Institute（伦斯勒理工学院）
We present a deep learning model for data-driven simulations of random dynamical systems without a distributional assumption. The deep learning model consists of a recurrent neural network, which aims to learn the time marching structure, and a generative adversarial network (GAN) to learn and sample from the probability distribution of the random dynamical system. Although GANs provide a powerful tool to model a complex probability distribution, the training often fails without a proper regularization. Here, we propose a regularization strategy for a GAN based on consistency conditions for th…
