Calibrating Energy-based Generative Adversarial Networks

By Zihang Dai, Amjad Almahairi, Philip Bachman, Eduard H. Hovy, Aaron C. Courville. 2017

111 citations recorded by the source corpus.

In this paper, we propose to equip Generative Adversarial Networks with the ability to produce direct energy estimates for samples. Specifically, we propose a flexible adversarial training framework, and prove this framework not only ensures the generator converges to the true data distribution, but also enables the discriminator to retain the density information at the global optimal. We derive the analytic form of the induced solution, and analyze the properties. In order to make the proposed framework trainable in practice, we introduce two effective approximation techniques. Empirically, the experiment results closely match our theoretical analysis, verifying the discriminator is able to recover the energy of data distribution.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Statistical and Nonlinear Physics, Generative Models

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Statistical and Nonlinear Physics

In Generative Models

Citations in the corpus

Flow Contrastive Estimation of Energy-Based Models (2019, background): “CD has been generalized to persistent CD [67], and has more recently been generalized to modiﬁed CD [13], adversarial CD [29, 5, 22] with modern CNN structure.”
Energy-Inspired Models: Learning with Sampler-Induced Distributions (2019, background): “Currently, most attempts to ﬁx the mismatch lie in designing better sampling algorithms ( e.g. , Hamiltonian Monte Carlo [53], annealed importance sampling [52]) or exploiting variational techniques [34, 15, 14] to reduce the inference approximation error.”
Learning Non-Convergent Non-Persistent Short-Run MCMC Toward Energy-Based Model (2019, methodology): “The maximum likelihood learning of the energy-based model (EBM) [29, 51, 19, 40, 30, 34, 6, 32, 48, 49, 22, 7, 47] follows what Grenander [15] called “analysis by synthesis” scheme.”
Learning Generative ConvNets via Multi-grid Modeling and Sampling (2017, background): “Another possibility is to recruit a generator network q α ( Y ) as an approximated direct sampler [16, 5], so that p θ and q α can be jointly learned by the adversarial CD: That is, the learning of p θ is modiﬁed CD with q α supplying synthesized examples, and the learning of q α is based on min q…”
Improved Training of Wasserstein GANs (2017, neutral): “06 EGAN-Ent-VI [7] 7 .”
Stacked Generative Adversarial Networks (2016, methodology): “Independent of our work, [4] proposes to regularize EBGAN [68] with entropy maximization in order to prevent the discriminator from degenerating to uniform prediction.”
LOGAN: Latent Optimisation for Generative Adversarial Networks (2019, methodology): “LOGAN is related to the energy-based formulation of a GAN’s discriminator (Dai et al., 2017; Kumar et al., 2019; Du & Mordatch, 2019), when latent optimisation is viewed as descending the energy function defined by the discriminator.”
Divergence Triangle for Joint Training of Generator Model, Energy-Based Model, and Inferential Model (2018, methodology): “Table 4 shows the IS and FID scores of our model compared with VAE [15], DCGAN [20], WGAN [22], CoopNet [25], CEGAN [24], ALI [26], ALICE [28].”
Maximum Entropy Generators for Energy-Based Models (2019, methodology): “Turning a GAN discriminator into an energy function has been studied in the past (Kim & Bengio, 2016; Zhao et al., 2016; Dai et al., 2017) but in order to turn a GAN discriminator into an energy function, a crucial and difﬁcult requirement is the maximization of entropy at the output of the…”
Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models (2020, methodology): “This includes variational training of energy based models (Dai et al., 2017; Zhai et al., 2016), entropy regularized policy gradient in reinforcement learning (Mazoure et al., 2019; Ward et al., 2019), probability distillation (Oord et al., 2018), and variational Bayesian inference of latent…”
Learning Neural Random Fields with Inclusive Auxiliary Generators (2018, background): “The recent work in [28] connects [27] and [8], and show another two approximations for the entropy term.”
MINE: Mutual Information Neural Estimation (2018, methodology): “Our approach is closest to that of Dai et al. (2017), where they also formulated a GAN with entropy regularization of the generator.”
Exponential Family Estimation via Adversarial Dynamics Embedding (2019, methodology): “…of the initial distribution is discussed in Appendix C. Compared to the optimization in GANs (Goodfellow et al., 2014; Arjovsky et al., 2017; Dai et al., 2017), beside the reversal of in (20), the major difference is that our “generator” (the dual sampler) shares parameters with the…”
On the Effectiveness of Least Squares Generative Adversarial Networks (2017, background): “[36] extended the energy-based GANs by adding some regularizations to make the discriminator non-degenerate.”