首页 > AI前沿 > Gaussian Flow-Matching Schedules: Implications for Sampling and Training

Gaussian Flow-Matching Schedules: Implications for Sampling and Training

arXiv机器学习 2026-09-22 16:06 3 阅读 查看原文

Flow-matching schedules affect both sampling dynamics and the variance of the regression target.

For centered commuting Gaussians, we show that a direction-dependent schedule decomposes into two independent design choices:

  • a variance path, which fully determines the intermediate laws and probability flow,
  • a factorization, which leaves this flow unchanged while controlling irreducible regression variance.

On the sampling side, we analyze finite-step Euler accuracy and derive a necessary drift bound for exact N -step sampling, connecting the geodesic and the logarithmic path.

On the training side, for any fixed path, we derive closed-form factorizations that either minimize time-averaged regression variance or make it constant along the path.