Importance Weighted Adversarial Nets for Partial Domain Adaptation

By Jing Zhang, Zewei Ding, Wanqing Li, Philip O. Ogunbona. 2018

450 citations recorded by the source corpus.

This paper proposes an importance weighted adversarial nets-based method for unsupervised domain adaptation, specific for partial domain adaptation where the target domain has less number of classes compared to the source domain. Previous domain adaptation methods generally assume the identical label spaces, such that reducing the distribution divergence leads to feasible knowledge transfer. However, such an assumption is no longer valid in a more realistic scenario that requires adaptation from a larger and more diverse source domain to a smaller target domain with less number of classes. This paper extends the adversarial nets-based domain adaptation and proposes a novel adversarial nets-based partial domain adaptation method to identify the source samples that are potentially from the outlier classes and, at the same time, reduce the shift of shared classes between domains.

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

A Comprehensive Survey on Transfer Learning (2019, background): “[162] proposes an approach for partial domain adaptation, which is called importance weighted adversarial nets-based domain adaptation (IWANDA).”
Universal Source-Free Domain Adaptation (2020, methodology): “[54] proposed to utilize importance weights for source samples depending on their similarity to the target domain data using an auxilliary discriminator.”
Strong-Weak Distribution Alignment for Adaptive Object Detection (2018, background): “In settings such as open-set classification [4, 33] or partial domain adaptation [41], strong alignment can be infeasible and could actually hurt performance.”
Domain-Symmetric Networks for Adversarial Domain Adaptation (2019, methodology): “Entropy minimization principle [7] is adopted by some domain adaptation methods [13, 33, 24] to enhance discrimination of learned models for target data.”
Overcoming Negative Transfer: A Survey (2020, background): “So the proposed weighted-discriminator loss is as follow: where Zhang et al. [43] further proposed a two domain clas-siﬁer strategy to identify the source samples that are potentially from the outlier classes and, at the same time, reduce the shift of shared classes between domains.”
A Survey on Negative Transfer (2020, background): “However, these assumptions may not always hold in many real-world applications [30], [102], [104].”
Multi-Domain Adversarial Learning (2019, background): “Notable exceptions are partial DA methods, where target classes are expected to be a subset of source classes e.g. (Zhang et al., 2018; Cao et al., 2018).”
Discriminative Partial Domain Adversarial Network (2020, methodology): “Methods are devised [1,2,32] to reduce label space’s mismatch by weighting each sample into the domain adversarial network.”
A Balanced and Uncertainty-aware Approach for Partial Domain Adaptation (2020, methodology): “ the source domain. Until recent years, [31] rst introduces an imbalanced scenario where label numbers of the source and target domains are not the same, which draws the attention of many researchers [38,4,5,49,30,6,17]. Dierent from shallow methods [31,38], deep methods [4,5,49,30] are mainly based”
Universal Domain Adaptation through Self Supervision (2020, background): “This task is solved by performing importance-weighting on source examples that are similar to samples in the target [2, 45, 3].”
A Survey of Unsupervised Deep Domain Adaptation (2018, background): “ce gains to other problems as well. Finally, another under-studied case of domain adaptation is partial domain adaptation, where the target domain contains only a subset of the source domain’s labels [26,233,272]. 9 CONCLUSIONS For supervised learning, deep neural networks are in prevalent use, but ”
Multi-Adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection (2019, methodology): “The straightforward way to solve this problem is to exploit the domain adaptation technique [27, 12, 25, 20, 34, 7, 24, 29, 6, 30, 33, 3] to align the feature distribution between training and testing data so that the trained model with source data can be adapted on the target data.”
Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation (2020, methodology): “For DA methods, (Zhang et al., 2018b; Choi et al., 2019) directly incorporate pseudo labeling as a regularization while (Long et al., 2017; 2018) leverage pseudo labels in the adaptation module to pursue discriminative (conditional) distribution alignment.”
Domain Adaptation by Joint Distribution Invariant Projections (2020, background): “By relaxing this assumption, the original problem can be morphed to new ones such as heterogeneous domain adaptation [18], open set domain adaptation [19], and partial domain adaptation [20].”
Deep Visual Domain Adaptation: A Survey (2018, background): “Recently, some papers [8], [7], [136] have begun to focus on this issue and we believe it is worthy of more attention.”