Generating more realistic images using gated MRF's

By Marc'Aurelio Ranzato, Volodymyr Mnih, Geoffrey E. Hinton. 2010

79 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Disentangling Factors of Variation via Generative Entangling (2012, background): “This is an important property of the spike-and-slab framework that is also shared by other latent variable models of real-valued data such as the meancovariance restricted Boltzmann machine (mcRBM) (Ranzato and Hinton, 2010) and the mean Product of T-distributions model (mPoT) (Ranzato et al., 2010)”
Learning generative models of mid-level structure in natural images (2012, methodology): “…image structure captured by the models (e.g. Olshausen and Field, 1997; Bell and Sejnowski, 1997; Lewicki and Olshausen, 1999; Hyvärinen and Hoyer, 2000; Roth and Black, 2005; Osindero and Hinton, 2008; Lee et al., 2009; Karklin and Lewicki, 2009; Sinz et al., 2010; Ranzato et al., 2010a,b).”
Learning Deep Energy Models (2011, methodology): “, 2003) and in generating large realistic samples (Ranzato et al., 2010).”
Unsupervised Models of Images by Spikeand-Slab RBMs (2011, methodology): “Qualitatively , these samples are more compelling than samples from similar energy-based models, such as those featured in Ranzato et al. (2010).”
Representation Learning: A Review and New Perspectives (2012, methodology): “The mPoT model (mean-product of Student’s T-distributions model) [164] is a combination of the GRBM and the product of Student’s T-distributions model [216].”
Attribute2Image: Conditional Image Generation from Visual Attributes (2015, background): “Early works [30,32,36,21,26,20] studied statistical and physical principles of building generative models, but due to the lack of effective feature representations, their results are limited to textures or particular patterns such as well-aligned faces.”
Hierarchical model of natural images and the origin of scale invariance (2013, background): “It may also be relevant in neuroscience because it suggests a neural architecture in the brain for generating images (6, 18).”
Visual Boundary Prediction: A Deep Neural Prediction Network and Quality Dissection (2014, methodology): “The mcRBM/mcDBN models were trained using stochastic gradient ascent for approximate maximum likelihood learning using FPCD [Tieleman and Hinton, 2009], with an implementation based on the TmPoT training code [Ranzato et al., 2010].”
Optimizing Neural Networks that Generate Iimages (2014, methodology): “The most obvious addition to the encoder would be some convolutional-style locality and weight sharing, or perhaps slightly less weight sharing as is done in (Ranzato et al., 2010).”
PriorGAN: Real Data Prior for Generative Adversarial Nets (2020, background): “Later works, such as the Hidden Markov Model (HMM) [29], Markov Random Field (MRF) [25], and restricted Boltzmann machines (RBMs) [27]. limiting their results on texture patches, digital numbers, due to a lack of effective feature representations.”
On deep generative models with applications to recognition (2011, methodology): “In previous work [18], we showed that mPoT (which is the first layer in our deep”
CKD: Cross-Task Knowledge Distillation for Text-to-Image Synthesis (2020, methodology): “Although the images synthesized by the early methods such as [22] are easy to be distinguished from real samples, the recent method [11], [23], [24] can generate photo-realistic images.”
On Autoencoders and Score Matching for Energy Based Models (2011, methodology): “Recently, more sophisticated latent variable EBMs for continuous data including the PoT (Welling et al., 2003), mPoT (Ranzato et al., 2010b), mcRBM (Ranzato & Hinton, 2010), FoE (Schmidt et al., 2010) and others have become popular models for learning representations of natural images as well as…”
CVAE-GAN: Fine-Grained Image Generation through Asymmetric Training (2017, background): “], all assume simple formation of the data. They have difﬁculty in modeling complex patterns of irregular distributions. Later works, such as Hidden Markov Model (HMM) [35], Markov Random Field (MRF) [19], restricted Boltzmann machines (RBMs) [9, 32], discriminatively trained generative models [39],”
Deep Facial Expression Recognition: A Survey (2018, methodology): “ppearance of the original facial expression, are two major obstacles for automatic FER, especially in real-world scenarios. Ranzato et al. [203], [204] proposed a deep generative model that used mPoT [205] as the ﬁrst layer of DBNs to model pixellevel representations and then trained DBNs to ﬁt an a”
Unsupervised Feature Learning and Deep Learning: A Review and New Perspectives (2012, methodology): “The mPoT model has been used to synthesize large-scale natural images (Ranzato et al. , 2010b) that show large-scale features and shadowing structure.”