Joint Geometrical and Statistical Alignment for Visual Domain Adaptation

By Jing Zhang, Wanqing Li, Philip O. Ogunbona. 2017

536 citations recorded by the source corpus.

This paper presents a novel unsupervised domain adaptation method for cross-domain visual recognition. We propose a unified framework that reduces the shift between domains both statistically and geometrically, referred to as Joint Geometrical and Statistical Alignment (JGSA). Specifically, we learn two coupled projections that project the source domain and target domain data into low-dimensional subspaces where the geometrical shift and distribution shift are reduced simultaneously. The objective function can be solved efficiently in a closed form. Extensive experiments have verified that the proposed method significantly outperforms several state-of-the-art domain adaptation methods on a synthetic dataset and three different real world cross-domain visual recognition tasks.

Subfields: Artificial Intelligence, Cancer Research, 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 Cancer Research

In Computer Vision and Pattern Recognition

Citations in the corpus

Self-Supervised Training Enhances Online Continual Learning (2021, methodology): “This set of experiments is similar to standard transfer learning and domain adaptation setups [42, 43, 52, 70], and could prove especially useful when performing continual learning on small datasets that require rich and generalizable feature representations.”
A Comprehensive Survey on Transfer Learning (2019, background): “Note that JDA can be extended by utilizing the label and structure information [79], clustering information [80], various statistical and geometrical information [81], and so on 4) μ ∈ (0, 1) and λ1 = 0: The article by Wang et al.”
A Survey on Deep Transfer Learning (2018, methodology): “New Data Space Transfer component analysis (TCA) introduced by [18] and TCA-based meth-ods [29] had been widely used in many applications of traditional transfer learning.”
Distant Supervised Centroid Shift: A Simple and Efficient Approach to Visual Domain Adaptation (2019, methodology): “Different from previous works [41, 74, 36] that generate pseudo target labels via merely exploiting the source data points, we aim to obtain a self-training target classifier and assume that the parameters are shifted from the parameters in the source classifier [71].”
Overcoming Negative Transfer: A Survey (2020, background): “In machine learning, TL has a rapid evolution during last decade, such as traditional transfer learning [3]–[6], deep transfer learning [7], [8] and adversarial transfer learning [9], [10], and it has been demonstrated to be effective in different applications, e.g., visual recognition, data…”
Unsupervised Multi-Target Domain Adaptation: An Information Theoretic Approach (2018, background): “Recently proposed unsupervised DA methods [27]–[30] operate by training deep neural networks using adversarial training, which allows the learning of feature representations that are simultaneously discriminative of source labels, and indistinguishable between the source and target domain.”
Learning Cross-Domain Disentangled Deep Representation with Supervision from A Single Domain (2017, background): “across domains to alleviate the domain bias, and Ghifary et al. [6] presented scatter component analysis to maximize the separability of classes and minimize the mismatch across domains. Zhang et al. [33] utilized coupled dimension reduction across data domains to reduce the geometrical and distribu”
Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift (2018, methodology): “Besides of these two mainstreams, there are diverse methods to learn domain-invariant features: semi-supervised method [42], domain reconstruction [14], duality [19], alignments [9] [50] [44], manifold learning [15], tensor methods [24][31], etc. Domain adaptation with multiple sources.”
Detach and Adapt: Learning Cross-Domain Disentangled Deep Representation (2017, methodology): “GFK [8] JDA [22] SA [4] TJM [23] SCA [6] JGSA [33] DC [30] GR [5] CoGAN [20] ADDA [31] DRCN [7] ADGAN [28] CDRD”
Domain Adaptation by Joint Distribution Invariant Projections (2020, background): “In general, most domain adaptation methods that rely on probability distribution matching attempt to either match the source and target marginal distributions [24]–[27], [29]– [32], or the source and target marginal distributions and class-conditional distributions [8], [11], [36]–[39], [43], [44].”
Visual Domain Adaptation with Manifold Embedded Distribution Alignment (2018, methodology): “oòpoooüpooùoôp (c) Target: Type II Figure1: Examplesoftwodi‡erenttargetdomainsw.r.t. the same source domain during distribution adaption. Unevaluateddistributionalignment means that existing work [19, 23, 33, 40] only a−empted to align the marginal and conditional distributions with equal weights. B”
A Survey on Negative Transfer (2020, background): “, traditional (statistical) TL [4]–[7], deep TL [8], [9], adversarial TL [10], [11], etc.”
Learning Compatible Embeddings (2021, unknown): “JDA is further extended by utilizing the label and structure information [18], clustering information [41], various statistical and geometrical information [53] and balanced distributions [47], etc.”
Domain-Adaptive Few-Shot Learning (2020, methodology): “The conventional UDA models [10, 15, 36, 13, 55, 56, 64, 29, 31] typically leverage the subspace alignment technique.”
Multicenter Computer-Aided Diagnosis for Lymph Nodes Using Unsupervised Domain-Adaptation Networks Based on Cross-Domain Confounding Representations (2020, methodology): “*ere are two types of methods in domain adaption: data-centric methods and subspace-centric methods [5].”
Importance Weighted Adversarial Nets for Partial Domain Adaptation (2018, methodology): “Since recent advance in deep learning has shown that more transferable and domain invariant features can be extracted through deep framework, the domain adaptation techniques are also transferred from shallow learningbased [1, 20, 15, 16, 8, 6, 28] to deep learning-based [25, 14, 18, 17, 27, 22, 23,”