用于条件流匹配的难度校准插值路径

中文摘要 AI 辅助

条件流匹配通过将网络回归到规定的噪声-数据插值路径的速度来训练生成模型。已知塑造该路径的插值调度会影响收敛性和样本质量，但它总是预先固定，与数据和模型均无关。我们表明，条件流匹配的回归难度沿路径呈系统性变化，因此提出了难度校准流匹配（Difficulty-Calibrated Flow Matching），该方法从模型本身推导调度：用线性路径进行的简短预运行记录了每个时间步的损失，调度被设置为该难度分布的分位数函数，从而使轨迹在最难学习速度的地方停留。该方法仅有一个超参数，保持训练目标及其梯度等价性不变，可与无分类器引导（classifier-free guidance）结合，且仅增加约2%的训练开销。在使用相同紧凑U-Net在CIFAR-10、MNIST和Fashion-MNIST上进行的对照实验中，校准路径在完整采样预算下于CIFAR-10上取得了最佳FID，且在计算资源最稀缺的大批次、少更新 regime 中明显优于所有固定调度。

英文摘要

Conditional Flow Matching trains generative models by regressing a network onto the velocity of a prescribed noise-to-data interpolation path. The interpolation schedule that shapes this path is known to affect convergence and sample quality, yet it is invariably fixed in advance, independent of both the data and the model. We show that the regression difficulty of Conditional Flow Matching varies systematically along the path, and we propose Difficulty-Calibrated Flow Matching, which derives the schedule from the model itself: a short pilot run with the linear path records the per-time loss, and the schedule is set to the quantile function of this difficulty profile, so the trajectory lingers where the velocity is hardest to learn. The method has a single hyperparameter, leaves the training objective and its gradient equivalence intact, composes with classifier-free guidance, and adds about two percent training overhead. In controlled experiments on CIFAR-10, MNIST, and Fashion-MNIST with an identical compact U-Net, the calibrated path attains the best FID on CIFAR-10 at full sampling budget and clearly outperforms all fixed schedules in the large-batch, few-update regime, precisely the setting where compute is scarcest.

发表机构

• Khulna University of Engineering & Technology（库尔纳工程技术大学）
• North Western University（西北大学）

机构由 AI 辅助整理，请以论文原文为准。