Latent common manifold learning with alternating diffusion: analysis and applications (Q2330938)

The article proposes a mathematical model based on alternating diffusion that is extended to more than two sensors. The model captures the nonlinear manifold structure by preserving both the geometrical and topological structures, it does not reply on prior knowledge and it can integrate subtle patterns of the data and even multiple data sets that contain complementary information. The efficiency of the proposed method is demonstrated and validated on several applications, from toy problems to complex tasks.