Generalization and Memorization: The Bias Potential Model
arXiv:2011.14269
Abstract
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 either memorizes the samples or diverges.
arXiv commentAdded new section on regularized model
References in corpus (20)
- Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
- Language Models are Few-Shot Learners
- Improved Techniques for Training GANs
- Understanding deep learning requires rethinking generalization
- A Mean Field View of the Landscape of Two-Layers Neural Networks
- NVAE: A Deep Hierarchical Variational Autoencoder
- A Variational Approach to Enhanced Sampling and Free Energy Calculations
- Mode Regularized Generative Adversarial Networks
- CAN: Creative Adversarial Networks, Generating "Art" by Learning About Styles and Deviating from Style Norms
- Neural networks-based variationally enhanced sampling
- A Priori Estimates of the Population Risk for Two-layer Neural Networks
- Density Estimation in Infinite Dimensional Exponential Families
- Towards a Mathematical Understanding of Neural Network-Based Machine Learning: what we know and what we don't
- A Priori Estimates of the Population Risk for Residual Networks
- Monge-Ampère Flow for Generative Modeling
- Approximability of Discriminators Implies Diversity in GANs
- Metrizing Weak Convergence with Maximum Mean Discrepancies
- On the Approximation Properties of Random ReLU Features
- Learning Distributions Generated by One-Layer ReLU Networks
- Estimating Densities with Non-Parametric Exponential Families