Implicit Generation and Modeling with Energy Based Models

By Yilun Du, Igor Mordatch. 2019

560 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

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

In Probabilistic & Bayesian Methods

Citations in the corpus

Denoising Diffusion Probabilistic Models (2020, background): “…sample quality, our models do not have competitive log likelihoods compared to other likelihood-based models (our models do, however, have log likelihoods better than the large estimates annealed importance sampling has been reported to produce for energy based models and score matching [11, 52]).”
Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling (2020, methodology): “Samples are typically generated from p ( x ) by an MCMC algorithm.”
Score-Based Generative Modeling through Stochastic Differential Equations (2020, neutral): “Score-based generative models, and related techniques (Bordes et al., 2017; Goyal et al., 2017; Du & Mordatch, 2019), have proven effective at generation of images (Song & Ermon, 2019; 2020; Ho et al., 2020), audio (Chen et al., 2020; Kong et al., 2020), graphs (Niu et al., 2020), and shapes (Cai ˚…”
VAEBM: A Symbiosis between Variational Autoencoders and Energy-based Models (2020, methodology): “There are 2 major differences between our energy networks and the ones used in Du & Mordatch (2019): 1. we replace the LeakyReLU activations with Swish activations, as we found it improves training stability, and 2. we do not use spectral normalization (Miyato et al., 2018); instead, we use weight…”
Generating Images with Sparse Representations (2021, background): “Energy-based models estimate a scalar energy for each example that corresponds to an unnormalized log-probability, and can be trained with a variety of objectives (Du & Mordatch, 2019).”
Energy-based Latent Aligner for Incremental Learning (2022, methodology): “Energy-based Models: EBMs [26] are a type of maximum likelihood estimation models that can assign low energies to observed data-label pairs and high energies otherwise [11].”
Autoregressive Denoising Diffusion Models for Multivariate Probabilistic Time Series Forecasting (2021, methodology): “In contrast EMBs do not suffer from these issues (Du & Mordatch, 2019).”
Energy-Based Models for Code Generation under Compilability Constraints (2021, background): “A recent surge of interest in EBMs (Du and Mordatch, 2019) has not left text generation unaffected (see (Bakhtin et al., 2020) for a survey).”
Training Deep Energy-Based Models with f-Divergence Minimization (2020, methodology): “Since the performance is sensitive to the model architectures, for fair comparisons, we use the same architecture and training hyper-parameters for f -EBMs and the contrastive divergence baseline (Du & Mordatch, 2019).”
NCP-VAE: Variational Autoencoders with Noise Contrastive Priors (2020, background): “Among them, energy-based models (EBMs) (Du & Mordatch, 2019; Pang et al., 2020) have shown promising results in learning expressive priors.”
Score-based diffusion models for accelerated MRI (2021, methodology): “This work grounds their method on an EBM, trained with persistent contrastive divergence (PCD), which is a variant of CD using sample buffers (Du and Mordatch, 2019).”
Energy-Based Imitation Learning (2020, background): “Among them, recent energy-based models (EBMs) have gained much attention because of the simplicity and flexibility in likelihood estimation [10, 41].”
COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics (2022, unknown): “Langevin dynamics is widely used on EBMs of modalities with continuous values, like images [48, 9, 59], 3D shapes [55], latent features [39], and audio sequences [22].”
Set Prediction without Imposing Structure as Conditional Density Estimation (2020, methodology): “…cutoff at T and missing Metropolis-Hastings update result in a biased sampler, previous works have demonstrated the feasibility of training generative models on images with similar Langevin MCMC methods (Xie et al., 2016; 2018; Nijkamp et al., 2020; Du & Mordatch, 2019; Grathwohl et al., 2019).”
Efficient Learning of Generative Models via Finite-Difference Score Matching (2020, background): “EBMs allow more flexible architectures [9, 11] with simpler compositionality [15, 41] compared to other explicit generative models [46, 16], and have better stability and mode coverage in training [30, 31, 71] compared to implicit generative models [10].”