学习直线流通过学习弯曲插值
Learning Straight Flows by Learning Curved Interpolants
AI总结:
本文提出通过学习弯曲插值来学习直线向量场,以加速流匹配模型的生成过程,并采用多级优化与高效近似实现端到端无模拟训练。
摘要:
流匹配模型通常使用线性插值来定义前向/噪声添加过程。这加上噪声与目标分布之间的独立耦合,产生了一个通常非直线的向量场。这种弯曲的场导致推理/生成过程缓慢。在这项工作中,我们提出学习灵活的(可能弯曲的)插值,以学习直线向量场,从而实现更快的生成。我们通过多级优化问题来形式化这一目标,并提出一个高效的近似过程来解决它。我们的框架提供了一个端到端且无模拟的优化过程,可用于学习直线生成轨迹。
英文摘要:
Flow matching models typically use linear interpolants to define the forward/noise addition process. This, together with the independent coupling between noise and target distributions, yields a vector field which is often non-straight. Such curved fields lead to a slow inference/generation process. In this work, we propose to learn flexible (potentially curved) interpolants in order to learn straight vector fields to enable faster generation. We formulate this via a multi-level optimization problem and propose an efficient approximate procedure to solve it. Our framework provides an end-to-end and simulation-free optimization procedure, which can be leveraged to learn straight line generative trajectories.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。