Multi-view Information Bottleneck Without Variational Approximation
This work addresses multi-view classification tasks, offering a more robust method for handling noise and redundancy, but it appears incremental as it builds on existing information bottleneck principles with a new optimization approach.
The authors tackled the problem of fusing complementary information across different views in multi-view learning for classification by extending the information bottleneck principle and using matrix-based Rényi's entropy to optimize the objective without variational approximation, resulting in improved robustness to noise and redundant information, especially with limited training samples.
By "intelligently" fusing the complementary information across different views, multi-view learning is able to improve the performance of classification tasks. In this work, we extend the information bottleneck principle to a supervised multi-view learning scenario and use the recently proposed matrix-based R{é}nyi's $α$-order entropy functional to optimize the resulting objective directly, without the necessity of variational approximation or adversarial training. Empirical results in both synthetic and real-world datasets suggest that our method enjoys improved robustness to noise and redundant information in each view, especially given limited training samples. Code is available at~\url{https://github.com/archy666/MEIB}.