Learning Generative Models via Discriminative Approaches

By Zhuowen Tu. 2007

124 citations recorded by the source corpus.

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

Generative Adversarial Nets (2014, methodology): “Previous work has also taken the approach of using a discriminative criterion to train a generative model [29, 13].”
Flow Contrastive Estimation of Energy-Based Models (2019, methodology): “Generalizing from [69], [27, 40, 43] developed an intro-spective parameter estimation method, where the EBM is discriminatively learned and composed of a sequence of discriminative models obtained through the learning process.”
Learning Energy-Based Spatial-Temporal Generative ConvNets for Dynamic Patterns (2019, methodology): “Our model in (2) directly corresponds to a classifier, specifically a spatial-temporal discriminative ConvNet, in the following sense [22], [51], [52], [53], [54], [55].”
On the Anatomy of MCMC-based Maximum Likelihood Learning of Energy-Based Models (2019, methodology): “The INN model [32] learns unnormalized densities in a discriminative framework.”
Learning Generative ConvNets via Multi-grid Modeling and Sampling (2017, background): “Recently, building on the early work of [43], [15, 22] have developed an introspective learning method to learn the energy-based model, where the energy function is discrim-inatively learned.”
Dual Contradistinctive Generative Autoencoder (2020, neutral): “Inspired by the literature on instance-level classiﬁcation [50], approximating likelihood by classiﬁcation [60], and contrastive learning [20, 65, 22], we propose to model instance-level ﬁdelity by contrastive loss (commonly referred to as InfoNCE loss) [62].”
Wasserstein Introspective Neural Networks (2017, methodology): “Motivated by the generative via discriminative learning (GDL) framework [44], one could try to learn the generative model for the given examples, p(x|y = +1), by a sequentially learned distribution of the pseudo-negative samples, pt(x|y = −1;Wt) which is abbreviated as pWt(x) where Wt includes all t”
Divergence Triangle for Joint Training of Generator Model, Energy-Based Model, and Inferential Model (2018, background): “Generalizing [13], [14] developed an introspective learning method where the energy function is discriminatively learned, and the energy-based model is both a generative model and a discriminative model.”
Cooperative Training of Descriptor and Generator Networks (2016, methodology): “See also earlier work by (Tu, 2007; Welling et al., 2002).”
CVAE-GAN: Fine-Grained Image Generation through Asymmetric Training (2017, background): “Later works, such as the Hidden Markov Model (HMM) [35], Markov Random Field (MRF) [19], and restricted Boltzmann machines (RBMs) [9, 32], discriminatively train generative models [39], limiting their results on texture patches, digital numbers or well aligned faces, due to a lack of effective featu”
Hybrid Discriminative-Generative Training via Contrastive Learning (2020, unknown): “Prior work also explicitly adopt discriminative models for generative modelling and show this iterative bootstrapping leads to improvement in classification, texture modeling, and other applications [50, 27, 21].”
Learning Energy-based Model with Flow-based Backbone by Neural Transport MCMC (2020, methodology): “This idea has been explored in noise contrastive estimation (NCE) [19, 14] and introspective neural networks (INN) [54, 25, 38], where the correction is obtained by discriminative learning.”
Learning Active Basis Model for Object Detection and Recognition (2010, background): “See also Tu (2007) for a related model.”
Semi-supervised Learning with GANs: Manifold Invariance with Improved Inference (2017, background): “4] [21], and the generator tries to generate samples that maximize that loss [39, 11].”
Introspective Neural Networks for Generative Modeling (2017, methodology): “We discuss the main formulation first, which bears some level of similarity to GDL [40] with the replacement of the boosting algorithm [11] by convolutional neural networks [27].”
Inducing wavelets into random fields via generative boosting (2016, unknown): “The features that we use are not binary, and our sampling and learning methods are different from [20].”