Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AGLP: A Graph Learning Perspective for Semi-supervised Domain Adaptation

Houcheng Su, Mengzhu Wang, Jiao Li, Nan Yin, Liang Yang, Li Shen

TL;DR

A graph learning perspective (AGLP) for semi-supervised domain adaptation is proposed which applies the graph convolutional network to the instance graph which allows structural information to propagate along the weighted graph edges and is the first work to model structural information in SSDA.

Abstract

In semi-supervised domain adaptation (SSDA), the model aims to leverage partially labeled target domain data along with a large amount of labeled source domain data to enhance its generalization capability for the target domain. A key advantage of SSDA is its ability to significantly reduce reliance on labeled data, thereby lowering the costs and time associated with data preparation. Most existing SSDA methods utilize information from domain labels and class labels but overlook the structural information of the data. To address this issue, this paper proposes a graph learning perspective (AGLP) for semi-supervised domain adaptation. We apply the graph convolutional network to the instance graph which allows structural information to propagate along the weighted graph edges. The proposed AGLP model has several advantages. First, to the best of our knowledge, this is the first work to model structural information in SSDA. Second, the proposed model can effectively learn domain-invariant and semantic representations, reducing domain discrepancies in SSDA. Extensive experimental results on multiple standard benchmarks demonstrate that the proposed AGLP algorithm outperforms state-of-the-art semi-supervised domain adaptation methods.

AGLP: A Graph Learning Perspective for Semi-supervised Domain Adaptation

TL;DR

A graph learning perspective (AGLP) for semi-supervised domain adaptation is proposed which applies the graph convolutional network to the instance graph which allows structural information to propagate along the weighted graph edges and is the first work to model structural information in SSDA.

Abstract

In semi-supervised domain adaptation (SSDA), the model aims to leverage partially labeled target domain data along with a large amount of labeled source domain data to enhance its generalization capability for the target domain. A key advantage of SSDA is its ability to significantly reduce reliance on labeled data, thereby lowering the costs and time associated with data preparation. Most existing SSDA methods utilize information from domain labels and class labels but overlook the structural information of the data. To address this issue, this paper proposes a graph learning perspective (AGLP) for semi-supervised domain adaptation. We apply the graph convolutional network to the instance graph which allows structural information to propagate along the weighted graph edges. The proposed AGLP model has several advantages. First, to the best of our knowledge, this is the first work to model structural information in SSDA. Second, the proposed model can effectively learn domain-invariant and semantic representations, reducing domain discrepancies in SSDA. Extensive experimental results on multiple standard benchmarks demonstrate that the proposed AGLP algorithm outperforms state-of-the-art semi-supervised domain adaptation methods.

Paper Structure

This paper contains 17 sections, 18 equations, 5 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Preliminaries
  4. Semi-Supervised Domain Adaptation
  5. Cross-Domain Adaptive Clustering (CDAC)
  6. Source Label Adaptation (SLA)
  7. Structure-aware Alignment
  8. Class Centroid Alignment
  9. Implementation Details
  10. Experiments
  11. Experiment Datasets
  12. Comparison Methods and Settings
  13. Comparative Experiments
  14. Further Performance Analysis
  15. Ablation Study
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Illustration of our AGLP. The data structure is constructed to build graph information.
  • Figure 2: Overall framework of our model.
  • Figure 3: In the domain adaptation experiment of Office-Home 3-Shot A→C, we conducted a parameter analysis by varying the output channels.
  • Figure 4: (a) illustrates the convergence behavior of the four loss functions in our model during the Office-Home 3-Shot A$\to$C domain adaptation experiment. (b) depicts the accuracy variations of the four models throughout the same experiment.
  • Figure 5: In the Office-Home 3-Shot A→C domain adaptation experiment, the confusion matrix and visualization analysis were computed by randomly selecting 10 classes from the dataset.