Cross Validation Framework to Choose amongst Models and Datasets for Transfer Learning

By Erheng Zhong, Wei Fan, Qiang Yang, Olivier Verscheure, Jiangtao Ren. 2010

157 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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 Computer Vision and Pattern Recognition

Citations in the corpus

Multi-Adversarial Domain Adaptation (2018, methodology): “For all baseline methods, we either follow their original model selection procedures, or conduct transfer cross-validation (Zhong et al. 2010) if their model selection strategies are not speciﬁed.”
Deep Transfer Learning with Joint Adaptation Networks (2016, methodology): “nd report the standard error of the classiﬁcation accuracies by different experiments of the same transfer task. We perform model selection by tuning hyper-parameters using transfer cross-validation (Zhong et al., 2010). For MMD-based methods and JAN, we adopt Gaussian kernel with bandwidth set to m”
Central Moment Discrepancy (CMD) for Domain-Invariant Representation Learning (2017, methodology): “Suitable tuning procedures are transfer learning speciﬁc cross-validation methods (Zhong et al., 2010).”
A Theoretical Analysis of Metric Hypothesis Transfer Learning (2015, background): “One perspective could be to consider notions of reverse validations as used in some transfer learning/domain adaptation tasks (Bruzzone & Marconcini, 2010; Zhong et al., 2010).”
Transfer Learning with Dynamic Adversarial Adaptation Network (2019, methodology): “Other Hyperparameters are tuned via transfer cross validation [44].”
A Survey of Unsupervised Deep Domain Adaptation (2018, background): “sed, in which case semi-supervised learning is recommended in Section6). One tuning method not requiring labeled target data is reverse validation [73], which is a variant of reverse cross validation [286]. For a set of hyperparameters, the reverse validation risk can be estimated by first splitting”
Unsupervised domain adaptation for medical imaging segmentation with self-ensembling (2018, methodology): “An alternative method to solve this issue is to use reverse cross-validation (Zhong et al., 2010), which was also used in (Ganin et al., 2016).”
Joint distribution optimal transportation for domain adaptation (2017, methodology): “All the methods have hyper-parameters, that are selected using the reverse cross-validation of Zhong and colleagues [31].”
Adaptation Based on Generalized Discrepancy (2019, methodology): “Whereas there are other methods such as transfer cross validation (Zhong et al., 2010) to select the parameters for our algorithm, these methods require the use of importance weighting which as shown in (Cortes et al., 2010) is not theoretically justified.”
Transitive Transfer Learning (2015, background): “The second one is feature based knowledge transfer [29], that transforms both source and target data into a common feature space where data follow similar distributions.”
Domain-Adversarial Training of Neural Networks (2015, methodology): “In the following experiments of Sections 5.1.3 and 5.1.4, we select the hyper-parameters of each algorithm by using a variant of reverse cross-validation approach proposed by Zhong et al. (2010), that we call reverse validation .”
Unsupervised Domain Adaptation with Similarity Learning (2017, methodology): “All hyperparameters were chosen via transfer crossvalidation [61].”
A New PAC-Bayesian Perspective on Domain Adaptation (2015, unknown): “We see this as a great source of inspiration for new algorithms for this learning paradigm.”