Balancing Complementarity and Consistency via Delayed Activation in Incomplete Multi-view Clustering

TL;DR

This work tackles incomplete multi-view clustering by preserving complementary information often neglected in traditional methods. It introduces CoCo-IMC, a framework that uses a delayed-activation dual network to balance complementarity and consistency across views, coupled with a reconstruction objective to recover missing information. The method optimizes a combined loss $\mathcal{L}= \beta \mathcal{L}_{cml} + \lambda \mathcal{L}_{cnl} + \alpha \mathcal{L}_{rec}$ and integrates cross-view mutual information and conditional entropy constraints alongside per-view autoencoders. Experiments on four public datasets against 12 baselines demonstrate state-of-the-art clustering performance under missing data and highlight the practical value of balancing complementary information with consistency in multi-view learning.

Abstract

This paper study one challenging issue in incomplete multi-view clustering, where valuable complementary information from other views is always ignored. To be specific, we propose a framework that effectively balances Complementarity and Consistency information in Incomplete Multi-view Clustering (CoCo-IMC). Specifically, we design a dual network of delayed activation, which achieves a balance of complementarity and consistency among different views. The delayed activation could enriches the complementarity information that was ignored during consistency learning. Then, we recover the incomplete information and enhance the consistency learning by minimizing the conditional entropy and maximizing the mutual information across different views. This could be the first theoretical attempt to incorporate delayed activation into incomplete data recovery and the balance of complementarity and consistency. We have proved the effectiveness of CoCo-IMC in extensive comparative experiments with 12 state-of-the-art baselines on four publicly available datasets.

Figures (9)

• Figure 1: Illustration of the motivation and the experimental results from multi-view data distribution on Caltech101-20. (a) Raw distribution. (b) Missing distribution. (c) Missing distribution with consistency. (d) Missing distribution with consistency and complementary. (e) Missing distribution with balancing consistency and complementary (Ours). In the figure, the green and blue curves denote the data distribution of two view, respectively. The orange curves denote the true distribution. We could observe that achieving a balance between complementarity and consistency in incomplete multi-view can closer to the true semantic, which is the most desirable result of multi-view clustering.
• Figure 2: Overview of CoCo-IMC. Bi-view data is used as a showcase in this figure. CoCo-IMC consists of three joint learning objectives, i.e., complementarity learning, consistency learning and reconstruction. Specifically, the complementary learning objective is to capture and balance the information of complementary and consistency. Consistency learning is to maximize the mutual information among different views. The goal of reconstruction is to project all views into a specific space.
• Figure 3: Complementary module.
• Figure 4: Consistency module.
• Figure 5: Parameter sensitivity analysis on Caltech101-20.