sinkhorn

Causal Neural Set Filtering (CNSF) is a proposed neural network method for online multi-target tracking that improves efficiency and performance by reducing error metrics by 19.3% and 30.4% compared to Track-MT3, with fewer parameters and faster inference.

This paper introduces MESH, a memory-efficient Sinkhorn-based optimizer for Mixture-of-Experts (MoE) training that restores temporal momentum without storing full optimizer state, reducing memory by 62.5% while maintaining competitive evaluation loss compared to AdamW.

This paper proposes a unified optimal transport framework for cold-start active learning, introducing a Sinkhorn-based algorithm (ε-AS) that adapts regularization strength to data and achieves state-of-the-art results on six datasets, including improving ImageNet-1k accuracy by 1.29% over prior methods while reducing selection time by 56.2%.