PhD student, University of Oulu
1 paper at NeurIPS 2025
We propose APML, a differentiable and efficient loss for point cloud prediction tasks, approximating one-to-one matching using Sinkhorn iterations with adaptive temperature.
