Stochastic Classifiers for Unsupervised Domain Adaptation

By Zhihe Lu, Yongxin Yang, Xiatian Zhu, Cong Liu, Yi-Zhe Song, Tao A. Xiang. 2020

178 citations recorded by the source corpus.

A common strategy adopted by existing state-of-the-art unsupervised domain adaptation (UDA) methods is to employ two classifiers to identify the misaligned local regions between source and target domain. Following the ’wisdom of the crowd’ principle, one has to ask: why stop at two? Indeed, we find that using more classifiers leads to better performance, but also introduces more model parameters, therefore risking overfitting. In this paper, we introduce a novel method called STochastic clAssifieRs (STAR) for addressing this problem. Instead of representing one classifier as a weight vector, STAR models it as a Gaussian distribution with its variance representing the inter-classifier discrepancy. With STAR, we can now sample an arbitrary number of classifiers from the distribution, whilst keeping the model size the same as having two classifiers. Extensive experiments demonstrate that a variety of existing UDA methods can greatly benefit from STAR and achieve the state-of-the-art performance on both image classification and semantic segmentation tasks.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

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

1. Jiashi Feng
2. Ziwei Liu
3. Yu Qiao
4. Chen Change Loy
5. Tao A. Xiang
6. Yi Yang
7. Kaiyang Zhou
8. Wen Li

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Jiashi Feng
3. Yi Yang
4. Trevor Darrell
5. Xiaogang Wang
6. Dacheng Tao
7. Chen Change Loy
8. Yu Qiao

In Radiology, Nuclear Medicine and Imaging

1. Jian Liang
2. Jiashi Feng
3. Joost van de Weijer
4. Ziwei Liu
5. Dapeng Hu
6. Soma Biswas
7. Jaya Kalla
8. Luis Herranz

Citations in the corpus

• S3C: Self-Supervised Stochastic Classifiers for Few-Shot Class-Incremental Learning (2023, methodology): “Unlike the standard classifiers, stochastic classifiers (SC) are represented by weight distributions, i.e. a mean and variance vector [24].”
• Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation (2021, background): “Additionally, [18, 26, 35] adopts prediction diversity between multiple learnable classifiers to achieve local or category-level feature alignment between source and target domains.”
• Denoised Maximum Classifier Discrepancy for Source-Free Unsupervised Domain Adaptation (2022, methodology): “In addition, there is also a special kind of adversarial method that depends on the adversarial training between the feature extractor and the classifier (Saito et al. 2018; Li et al. 2021; Lee et al. 2019; Lu et al. 2020).”
• Domain Generalization: A Survey (2021, background): “Indeed, this domain adaptation (DA) problem has received much attention [12], [13], [14], [15], [16], [17], [18].”
• Domain Consensus Clustering for Universal Domain Adaptation (2021, methodology): “Some methods minimize the discrepancy in the feature space directly [40, 39, 56, 59, 42, 61, 30, 16].”
• Source Data-Absent Unsupervised Domain Adaptation Through Hypothesis Transfer and Labeling Transfer (2020, result): “For vanilla unsupervisedDA in digit recognition, we compare SHOT with ADDA [5], ADR [56], CDAN [42], CyCADA [8], CAT [98], SWD [99] and STAR [100]; for object recognition, we compare ours with DANN [24], DAN [4], SAFN [82], BSP [101], MDD [102], TransNorm [55], DSBN [25], BNM [103] and GVB-GD [104].”
• FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation (2020, unknown): “6 STAR [24] 95 84 84.”