Learning to Discover Knowledge: A Weakly-Supervised Partial Domain Adaptation Approach

TL;DR

Self-paced transfer classifier learning (SP-TCL) learns to discover faithful knowledge via a carefully designed prudent loss function and simultaneously adapts the learned knowledge to the target domain by iteratively excluding source examples from training under the self-paced fashion.

Abstract

Domain adaptation has shown appealing performance by leveraging knowledge from a source domain with rich annotations. However, for a specific target task, it is cumbersome to collect related and high-quality source domains. In real-world scenarios, large-scale datasets corrupted with noisy labels are easy to collect, stimulating a great demand for automatic recognition in a generalized setting, i.e., weakly-supervised partial domain adaptation (WS-PDA), which transfers a classifier from a large source domain with noises in labels to a small unlabeled target domain. As such, the key issues of WS-PDA are: 1) how to sufficiently discover the knowledge from the noisy labeled source domain and the unlabeled target domain, and 2) how to successfully adapt the knowledge across domains. In this paper, we propose a simple yet effective domain adaptation approach, termed as self-paced transfer classifier learning (SP-TCL), to address the above issues, which could be regarded as a well-performing baseline for several generalized domain adaptation tasks. The proposed model is established upon the self-paced learning scheme, seeking a preferable classifier for the target domain. Specifically, SP-TCL learns to discover faithful knowledge via a carefully designed prudent loss function and simultaneously adapts the learned knowledge to the target domain by iteratively excluding source examples from training under the self-paced fashion. Extensive evaluations on several benchmark datasets demonstrate that SP-TCL significantly outperforms state-of-the-art approaches on several generalized domain adaptation tasks.

Figures (10)

• Figure 1: The problem of weakly-supervised partial domain adaptation, where the target label space is a subspace of the source label space and the source domain is naturally corrupted with noisy labels. For each noisy or incorrectly classified sample, the face-color denotes its ground truth, while the edge-color denotes its noisy label. For existing traditional DA methods, outlier classes will cause negative transfer, and noisy samples will affect the final classifier learning. In our SP-TCL, we propose knowledge discovery and knowledge adaptation strategies to effectively address the problem of weakly-supervised partial domain adaptation.
• Figure 2: The overview of generalized knowledge adaptation: (a) the original joint classifier, (b) self-pace learning excludes source data from training, (c) manifold learning enforces the decision boundary to locate in the low-density area of target data, resulting in target-preferable classifier.
• Figure 3: Image examples of Office31, ImageCLEF and Office-Home datasets.
• Figure 4: (a) Accuracies versus iterations on the task Cl$\to$Rw ; (b) Histograms of class probability distribution of Rw in the target domain on Cl$\to$Rw. The top row and bottom row show the results under the scenarios of PDA and UDA, respectively.
• Figure 5: Comparison of SP-TCL and its variant SP-TCL (w/o SPL) on (a) A$\to$D and (b) C$\to$P.