Deep Directed Generative Models with Energy-Based Probability Estimation

By Taesup Kim, Yoshua Bengio. 2016

141 citations recorded by the source corpus.

Training energy-based probabilistic models is confronted with apparently intractable sums, whose Monte Carlo estimation requires sampling from the estimated probability distribution in the inner loop of training. This can be approximately achieved by Markov chain Monte Carlo methods, but may still face a formidable obstacle that is the difficulty of mixing between modes with sharp concentrations of probability. Whereas an MCMC process is usually derived from a given energy function based on mathematical considerations and requires an arbitrarily long time to obtain good and varied samples, we propose to train a deep directed generative model (not a Markov chain) so that its sampling distribution approximately matches the energy function that is being trained. Inspired by generative adversarial networks, the proposed framework involves training of two models that represent dual views of the estimated probability distribution: the energy function (mapping an input configuration to a scalar energy value) and the generator (mapping a noise vector to a generated configuration), both represented by deep neural networks.

Subfields: Computer Vision and Pattern Recognition, Generative Models, Signal Processing

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 Signal Processing

Citations in the corpus

Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling (2020, methodology): “The contribution to training from the partition function can be estimated via MCMC [31, 34, 35], via training another generator network [36, 37], or via surrogate objectives to maximum likelihood [38–40].”
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.”
On the Anatomy of MCMC-based Maximum Likelihood Learning of Energy-Based Models (2019, background): “Learning a ConvNet potential with the help of a generator network as approximative direct sampler is explored in [18, 5, 36, 37, 11, 20].”
Learning Generative ConvNets via Multi-grid Modeling and Sampling (2017, methodology): “Within the same setting, we compare the learned features of the multi-grid method with the single-grid method, persistent CD [42], one-step CD [10], Wasserstein GAN [1], deep directed generative models [16] and DCGAN [37].”
Reinforcement Learning with Deep Energy-Based Policies (2017, methodology): “…generally fall into two categories: methods that use Markov chain Monte Carlo (MCMC) based sampling (Sallans & Hinton, 2004), and methods that learn a stochastic sampling network trained to output approximate samples from the target distribution (Zhao et al., 2016; Kim & Bengio, 2016).”
A Connection between Generative Adversarial Networks, Inverse Reinforcement Learning, and Energy-Based Models (2016, methodology): “Kim & Bengio proposed a similar energy-based model for gener ative image modeling, but did not assume they could compute the generator’s density [12].”
Calibrating Energy-based Generative Adversarial Networks (2017, methodology): “This elegant result has deep connections to several existing formulations, which include max-entropy imitation learning (Ziebart et al., 2008) and the directed-generator-trained energy-based model (Kim & Bengio, 2016).”
A Distributional Approach to Controlled Text Generation (2020, methodology): “On line 3, we then use SNIS (Self Normalized Importance Sampling) (Kim & Bengio, 2016; Parshakova et al., 2019a) to estimate µ(λ) .= Ex∼pφ(x).”
Learning Energy-Based Models by Diffusion Recovery Likelihood (2020, background): “EBMs [27, 33, 21, 47, 10, 43, 6, 7, 25, 35, 5, 11, 4, 8, 2, 12] are an appealing class of probabilistic models, which can be viewed as generative versions of discriminators [18, 26, 28, 12], yet can be learned from unlabeled data.”
Learning to Draw Samples: With Application to Amortized MLE for Generative Adversarial Learning (2016, methodology): “…methods based on MCMC-MLE or contrastive divergence, we amortize the sampler as we train , which gives much faster speed and simultaneously provides a high quality generative neural network that can generate realistic-looking images; see Kim & Bengio (2016) for a similar idea and discussions.”
Divergence Triangle for Joint Training of Generator Model, Energy-Based Model, and Inferential Model (2018, methodology): “Inspired by [38], we call (15) the adversarial contrastive divergence (ACD).”
Implicit Generation and Generalization in Energy-Based Models (2019, background): “(Kim and Bengio, 2016; Zhao et al., 2016; Haarnoja et al., 2017) pro-pose learning a separate network to generate samples, which makes these methods closely connected to generative adversarial networks as shown by (Finn et al., 2016), but these methods do not have the advantages of implicit…”
Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching (2018, methodology): “(Kim and Bengio 2016) learned the generator model by gradient descent on KL(qα|pθ(t)) over α.”
Maximum Entropy Generators for Energy-Based Models (2019, methodology): “The authors would also like to mention that the objective of the proposed method is not to beat the best GANs (which do not provide an energy function) but to show that it is possible to have both an energy function and good samples by appropriately ﬁxing issues from (Kim & Bengio, 2016).”
Learning Deep Energy Models: Contrastive Divergence vs. Amortized MLE (2017, unknown): “The idea of training energy models with neural samplers was also discussed by Kim & Bengio (2016); Zhai et al. (2016); one of the key differences is that the neural samplers in Kim & Bengio (2016); Zhai et al. (2016) are trained with the help of heuristic diversity regu-larizers, while SVGD…”