Classification with Hybrid Generative/Discriminative Models

By Rajat Raina, Yirong Shen, Andrew Y. Ng, Andrew McCallum. 2003

254 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.”
Supervised Contrastive Learning for Pre-trained Language Model Fine-tuning (2020, background): “Raina et al. (2003) show that a hybrid discriminative and generative objective outperforms both solely discriminative and generative approaches.”
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).”
Deep Clustering via Joint Convolutional Autoencoder Embedding and Relative Entropy Minimization (2017, background): “Generally it has been argued that the discriminative models often have better results compared to their generative counterparts, since they have fewer assumptions about the data distribution and directly separate the clusters, but their training can suffer from overfitting or getting stuck in undesi”
Discriminatively trained Markov model for sequence classification (2005, methodology): “Raina et. al. [ 18 ] proposed a model that combine s Na¨ ive Bayes Multinomial model and Logistic Regression, and showed that their hybrid model has lower error rate than either Na¨ ive Bayes or Logistic Regression.”
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].”
Spam Deobfuscation using a Hidden Markov Model (2005, methodology): “the out-of-dictionary HMM using the method given in [Raina et al., 2003] (essentially logistic regression).”
Hybrid Models with Deep and Invertible Features (2019, methodology): “…& Haussler (1999) were perhaps the ﬁrst to meaningfully combine the two, using a generative model to deﬁne a kernel function that could then be employed by classiﬁers such as SVMs. Raina et al. (2004) However, these models, unlike ours, require approximate inference to obtain the p ( x ) component.”
Towards 3D object recognition via classification of arbitrary object tracks (2011, methodology): “Second, we adapt the ideas of hybrid generative / discriminative models [1] to alleviate the overly strong conditional independence assumptions made when combining log odds estimates over time.”
Answer-Type Prediction for Visual Question Answering (2016, background): “Hybrid models that combine generative and discriminative classifiers can achieve a lower error rate than either alone [14].”
Hybrid Discriminative-Generative Training via Contrastive Learning (2020, background): “Later it was shown that hybrid discriminative generative model training can get the best of both worlds [40].”
Supervised Dictionary Learning (2008, background): “, [12, 13]), where generative components are often added to discriminative frameworks to add robustness, e.”
Semi-Supervised Structured Output Learning Based on a Hybrid Generative and Discriminative Approach (2007, methodology): “A hybrid approach was ﬁrst proposed in a supervised learning setting (Raina et al., 2003) for text classiﬁcation.”
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].”