The Tradeoff Between Generative and Discriminative Classifiers

By Guillaume Bouchard, Bill Triggs. 2004

170 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

Dropout Training as Adaptive Regularization (2013, background): “Most approaches to semi-supervised learning either rely on using a generative model [19, 20, 21, 22, 23] or various assumptions on the relationship between the predictor and the marginal distribution over inputs.”
Recent Advances in Open Set Recognition: A Survey (2018, background): “There is an ongoing debate between the use of generative and discriminative models in statistical learning [65], [66], with arguments for the value of each.”
Principled Hybrids of Generative and Discriminative Models (2006, methodology): “Recently several authors [2, 3] have proposed hybrids of the generative and discriminative approaches in which a model is trained by optimizing a convex combination of the generative and discriminative log likelihood functions.”
Discriminatively trained Markov model for sequence classification (2005, methodology): “For example, Ng and Jordan [15] have compared Naı̈ve Bayes and Logistic regression classifiers; Bouchard and Triggs [5] have explored the trade-offs between generative and discriminative classifiers; Raina et. al.”
Exploring Strategies for Training Deep Neural Networks (2009, background): “Moreover, successful attempts at exploring the space between discriminative and generative learning have been studied (Lasserre et al., 2006; Jebara, 2003; Bouchard and Triggs, 2004; Holub and Perona, 2005).”
Altitude Training: Strong Bounds for Single-Layer Dropout (2014, result): “Our analysis presents an interesting contrast to other work that directly combine generative and discriminative modeling by optimizing a hybrid likelihood [20, 21, 22, 23, 24, 25].”
Learning Algorithms for the Classification Restricted Boltzmann Machine (2012, methodology): “Since we might want to adapt the amount of regularization to the problem at hand, we could consider interpolating between the generative and discriminative objectives as in Bouchard and Triggs (2004) or, similarly, use the following hybrid objective:”
A Deep and Autoregressive Approach for Topic Modeling of Multimodal Data (2014, background): “The first term is a purely discriminative term, while the second is unsupervised and can be understood as a regularizer, that encourages a solution which also explains the unsupervised statistical structure within the visual words.”
Data Augmentation via Levy Processes (2016, background): “A number of papers study hybrid loss functions that combine both a joint and conditional likelihood (Raina et al., 2004; Bouchard and Triggs, 2004; Lasserre et al., 2006; McCallum et al., 2006; Liang and Jordan, 2008).”
Asymptotically Optimal Regularization in Smooth Parametric Models (2009, methodology): “We might regularize with a generative model if we think it is roughly well-specified [7, 20, 15, 17].”
Probability Models for Open Set Recognition (2014, background): “There is an ongoing debate between the use of generative and discriminative models in statistical learning [2], [20], with arguments for the value of each.”
Topic Modeling of Multimodal Data: An Autoregressive Approach (2014, unknown): “This is known as generative learning [18].”