Bias and Generalization in Deep Generative Models: An Empirical Study

By Shengjia Zhao, Hongyu Ren, Arianna Yuan, Jiaming Song, Noah D. Goodman, Stefano Ermon. 2018

160 citations recorded by the source corpus.

In high dimensional settings, density estimation algorithms rely crucially on their inductive bias. Despite recent empirical success, the inductive bias of deep generative models is not well understood. In this paper we propose a framework to systematically investigate bias and generalization in deep generative models of images. Inspired by experimental methods from cognitive psychology, we probe each learning algorithm with carefully designed training datasets to characterize when and how existing models generate novel attributes and their combinations. We identify similarities to human psychology and verify that these patterns are consistent across commonly used models and architectures.

Subfields: Computer Vision and Pattern Recognition, Generative Models, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Generative Models

In Evaluation & Benchmarks

Citations in the corpus

Denoising Diffusion Implicit Models (2020, background): “However, GANs require very speciﬁc choices in optimization and architectures in order to stabilize training (Arjovsky et al., 2017; Gulrajani et al., 2017; Karras et al., 2018; Brock et al., 2018), and could fail to cover modes of the data distribution (Zhao et al., 2018).”
Imperfect ImaGANation: Implications of GANs Exacerbating Biases on Facial Data Augmentation and Snapchat Selfie Lenses (2020, background): “G collapses because the set of noise inputs that There are several related works– Zhao et al. (2018) studies GANs’ bias and generalization to unseen modes without discussing the problem of GANs collapsing to existing modes.”
Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting (2019, methodology): “Also observe that the distribution p under which we seek expectations is same as p data here, and we rely on the generalization of p θ to generate transformations of an instance in the dataset which are not explicitly present, but plausibly observed in other, similar instances [21].”
Inclusive GAN: Improving Data and Minority Coverage in Generative Models (2020, background): “The biases are rooted in the inevitable imbalance in the dataset [42], which are preserved or even exacerbated by the generative models [57].”
Understanding the Limitations of Variational Mutual Information Estimators (2019, background): “…(MI) estimation and optimization are crucial to many important problems in machine learning, such as representation learning (Chen et al., 2016; Zhao et al., 2018b; Tishby & Zaslavsky, 2015; Higgins et al., 2018) and reinforcement learning (Pathak et al., 2017; van den Oord et al., 2018).”
Negative Data Augmentation (2021, background): “For example, Zhao et al. (2018) report that when trained on images containing 2 objects only, GANs and other generative models can some-times “generalize” by generating images with 1 or 3 objects (which were never seen in the training set).”
Compositionality and Generalization In Emergent Languages (2020, background): “inductive biases (Zhao et al., 2018).”
Tackling the Generative Learning Trilemma with Denoising Diffusion GANs (2021, background): “Generative adversarial networks (GANs) (Goodfellow et al., 2014; Brock et al., 2018) generate high-quality samples rapidly, but they have poor mode coverage (Salimans et al., 2016; Zhao et al., 2018).”