Zhaozhao Ma, Shujian Yu
This paper introduces a new framework for multimodal regression that enhances interpretability by decomposing modality contributions using Partial Information Decomposition (PID), achieving better accuracy and insight than existing methods.
Predicting outcomes using data from multiple sources can be challenging, especially when trying to understand the contribution of each data source. This research presents a new method that breaks down the information from each source into unique, shared, and combined effects using a mathematical tool called Partial Information Decomposition. By doing so, it provides clearer insights into how different data sources contribute to the prediction, improving both the accuracy and the interpretability of the results. The method was tested on several real-world datasets, including brain imaging data, and showed superior performance compared to existing techniques.
