Analysis of Representations for Domain Adaptation

By Shai Ben-David, John Blitzer, Koby Crammer, Fernando C Pereira. 2006

2,573 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning

How researchers see this paper

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

In Artificial Intelligence

In Representation & Self-Supervised Learning

Citations in the corpus

A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference (2017, background): “forms well on test data from a target domain with a different distribution—have resulted in gains across many tasks (Daume III and Marcu, 2006; Ben-David et al., 2007), including sequence and part-of-speech tagging (Blitzer et al.”
Domain-Adversarial Training of Neural Networks (2015, methodology): “The theoretical foundation of the DANN algorithm is the domain adaptation theory of Ben-David et al. (2006, 2010).”
Unsupervised Cross-Domain Image Generation (2016, methodology): “(2016) linked the GAN loss to the H-divergence between two distributions of Ben-david et al. (2006). The generative architecture that we employ is based on the successful architecture of Radford et al. (2015). There has recently been a growing concern about the uneven distribution of the samples gen”
Learning Transferable Features with Deep Adaptation Networks (2015, methodology): “We provide an analysis of the expected target-domain risk of our approach, making use of the theory of domain transfer (Ben-David et al., 2007; 2010; Mansour et al., 2009) and the theory of kernel embedding of probability distributions (Sriperumbudur et al.”
A theory of learning from different domains (2010, methodology): “Preliminary versions of the work contained in this article appeared in Advances in Neural Information Processing Systems ( Ben-David et al. 2006; Blitzer et al. 2007a).”
Marginalized Denoising Autoencoders for Domain Adaptation (2012, methodology): “Instead of learning a single mapping W ∈ R d × ( d +1) to reconstruct all corrupted features, we learn multiple mappings but only reconstruct the r (cid:28) d (Blitzer et. al., 2007) CODA (Chen et. al., 2011) SDA (Glorot et. al., 2011) mSDA (l=5) most frequent features (here, r = 5000 ).”
Multiple-Source Adaptation with Domain Classifiers (2020, background): “…are available for some source domains such as TVs, laptops and CD players, each previously trained on labeled data, but no labeled data or predictor is at hand for the broader category of electronics, which can be viewed as a mixture of the sub-domains [Blitzer et al., 2007, Dredze et al., 2008].”
A Theory of Multiple-Source Adaptation with Limited Target Labeled Data (2020, methodology): “lassiﬁ- cation task at hand. To obtain better bounds, we use the notion of discrepancy between distributions (Mansour et al.,2009c;Mohri and Medina,2012), which generalizes the dA distance proposed byBen-David et al. (2007). For two distributions over features and labels, D and D′, a class of hypoth”
Adaptive Personalized Federated Learning (2020, background): “As pointed out, related works by Ben-David et al. (2010, 2007); Crammer et al. (2008) did the pioneering analysis of learning a single model from diﬀerent domains.”
The Variational Fair Autoencoder (2015, methodology): “Briefly, Proxy Adistance is an approximation to the H-divergence measure of domain distinguishability proposed in Kifer et al. (2004) and Ben-David et al. (2007; 2010).”
Conditional variance penalties and domain shift robustness (2017, methodology): “(2016), an approach motivated by the work of Ben-David et al. (2007). The main idea of Ganin et al.”
Domain Adaptation with Conditional Transferable Components (2016, methodology): “Current analysis methods on domain adaptation (Ben-David et al., 2007; 2010) decompose the joint distribution P (X,Y ) to P (X)P (Y |X) and measure their distance between domains separately.”
Multi-Source Domain Adaptation: A Causal View (2015, background): “We remark that for some real problems, certain features could be highly noisy, and it is worth noting that this assumption might not hold for some features or components of features; therefore it would be beneficial to find appropriate feature representations, as in (Ben-David et al. 2007).”
Unsupervised Domain Adaptation with Residual Transfer Networks (2016, background): “domain discrepancy (Glorot et al., 2011; Tzeng et al., 2014). Dataset shift has posed a bottleneck to the transferability of deep features, resulting in statistically unbounded risk for target tasks (Ben-David et al., 2007; Mansour et al., 2009; Ben-David et al., 2010). Some recent work addresses th”
Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations (2020, background): “Empirically, minimizing theH−divergence involves building a representation where examples from the multiple domains are as indistinguishable as possible (Ben-David et al., 2007; Li et al., 2018; Sebag et al., 2019).”