Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic Evidence Decoupling for Trusted Multi-view Learning

Ying Liu, Lihong Liu, Cai Xu, Xiangyu Song, Ziyu Guan, Wei Zhao

TL;DR

This work proposes a Consistent and Complementary-aware trusted Multi-view Learning (CCML) method, which first construct view opinions using evidential deep neural networks, and dynamically decouple the consistent and complementary evidence.

Abstract

Multi-view learning methods often focus on improving decision accuracy, while neglecting the decision uncertainty, limiting their suitability for safety-critical applications. To mitigate this, researchers propose trusted multi-view learning methods that estimate classification probabilities and uncertainty by learning the class distributions for each instance. However, these methods assume that the data from each view can effectively differentiate all categories, ignoring the semantic vagueness phenomenon in real-world multi-view data. Our findings demonstrate that this phenomenon significantly suppresses the learning of view-specific evidence in existing methods. We propose a Consistent and Complementary-aware trusted Multi-view Learning (CCML) method to solve this problem. We first construct view opinions using evidential deep neural networks, which consist of belief mass vectors and uncertainty estimates. Next, we dynamically decouple the consistent and complementary evidence. The consistent evidence is derived from the shared portions across all views, while the complementary evidence is obtained by averaging the differing portions across all views. We ensure that the opinion constructed from the consistent evidence strictly aligns with the ground-truth category. For the opinion constructed from the complementary evidence, we allow it for potential vagueness in the evidence. We compare CCML with state-of-the-art baselines on one synthetic and six real-world datasets. The results validate the effectiveness of the dynamic evidence decoupling strategy and show that CCML significantly outperforms baselines on accuracy and reliability. The code is released at https://github.com/Lihong-Liu/CCML.

Dynamic Evidence Decoupling for Trusted Multi-view Learning

TL;DR

This work proposes a Consistent and Complementary-aware trusted Multi-view Learning (CCML) method, which first construct view opinions using evidential deep neural networks, and dynamically decouple the consistent and complementary evidence.

Abstract

Multi-view learning methods often focus on improving decision accuracy, while neglecting the decision uncertainty, limiting their suitability for safety-critical applications. To mitigate this, researchers propose trusted multi-view learning methods that estimate classification probabilities and uncertainty by learning the class distributions for each instance. However, these methods assume that the data from each view can effectively differentiate all categories, ignoring the semantic vagueness phenomenon in real-world multi-view data. Our findings demonstrate that this phenomenon significantly suppresses the learning of view-specific evidence in existing methods. We propose a Consistent and Complementary-aware trusted Multi-view Learning (CCML) method to solve this problem. We first construct view opinions using evidential deep neural networks, which consist of belief mass vectors and uncertainty estimates. Next, we dynamically decouple the consistent and complementary evidence. The consistent evidence is derived from the shared portions across all views, while the complementary evidence is obtained by averaging the differing portions across all views. We ensure that the opinion constructed from the consistent evidence strictly aligns with the ground-truth category. For the opinion constructed from the complementary evidence, we allow it for potential vagueness in the evidence. We compare CCML with state-of-the-art baselines on one synthetic and six real-world datasets. The results validate the effectiveness of the dynamic evidence decoupling strategy and show that CCML significantly outperforms baselines on accuracy and reliability. The code is released at https://github.com/Lihong-Liu/CCML.
Paper Structure (30 sections, 9 equations, 6 figures, 4 tables)

This paper contains 30 sections, 9 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multi-view Fusion
  4. Uncertainty-aware Deep Learning
  5. The Method
  6. Notations and Problem Definition
  7. CCML Pipeline
  8. View-specific Evidence Learning
  9. Consistent and Complementary-aware Multi-view Fusion
  10. Loss Function
  11. View-specific Loss Function
  12. Consistent Loss Function
  13. Separability Principle
  14. Error Reduction Principle
  15. Complementary Loss Function
  16. ...and 15 more sections

Figures (6)

  • Figure 1: Visualization of the dynamic semantic vagueness phenomenon. The ground-truth category of the first instance is "apple pie". When considering the image view alone, it becomes challenging to differentiate between the categories "waffles" and "apple pie". For the second instance, both views provide explicit differentiation between the categories.
  • Figure 2: Illustration of CCML. CCML initially constructs view-specific evidential DNNs to acquire the view-specific evidence and subsequently dynamically decouples the consistent and complementary evidence. During training, CCML ensures precise alignment between the opinion constructed from the consistent evidence and the ground-truth category. Regarding the complementary evidence, CCML only necessitates reflecting the probability of the true category, accommodating potential evidence vagueness. During testing, CCML combines consistent and complementary evidence to reach a decision.
  • Figure 3: Visualization of data instances in the toy dataset.
  • Figure 4: Visualization of the evidences (outputs of TMC and CCML on the toy dataset.
  • Figure 5: The accuracy with different hyper-parameter $\beta$.
  • ...and 1 more figures

Theorems & Definitions (2)

  • definition 1
  • definition 2