Solving Inverse Problems of Chaotic Systems with Bidirectional Conditional Flow Matching

arXiv·low signal

Introduces bidirectional conditional flow matching to solve inverse problems in chaotic dynamics — inferring initial conditions from final states, a long-standing hard problem. A specialized scientific-ML method with novel methodology but limited mainstream applicability.