UTV decomposition of dual matrices and its applications

Matrix factorization in the context of dual numbers has found applications, in recent years, in fields such as kinematics and computer graphics. In this paper, we develop an efficient approach for handling large-scale data low-rank approximation problems using the UTV decomposition of dual matrices (DUTV). Theoretically, we propose an explicit expression for the DUTV and provide necessary and sufficient conditions for its existence. During this process, we also discovered that the general low-rank model for dual matrices can be solved by the Sylvester equation. In numerical experiments, the DUTV algorithm outperforms the dual matrix SVD algorithm in terms of speed and maintains effective performance in wave recognition. Subsequently, we utilize the DUTV algorithm to validate brain functional circuits in large-scale task-state functional magnetic resonance imaging data. Successfully identifying three types of wave signals, the DUTV method provides substantial empirical evidence for cognitive neuroscience theories.