Rectified Noise: A Generative Model Using Positive-incentive Noise
This work addresses the need for more efficient and effective generative models in machine learning, though it appears incremental as it builds upon existing Rectified Flow methods.
The paper tackles the problem of improving generative model performance by proposing Rectified Noise (RN), a method that injects Positive-incentive Noise into pre-trained Rectified Flow models, resulting in reduced FID from 10.16 to 9.05 on ImageNet-1k with only 0.39% additional training parameters.
Rectified Flow (RF) has been widely used as an effective generative model. Although RF is primarily based on probability flow Ordinary Differential Equations (ODE), recent studies have shown that injecting noise through reverse-time Stochastic Differential Equations (SDE) for sampling can achieve superior generative performance. Inspired by Positive-incentive Noise (pi-noise), we propose an innovative generative algorithm to train pi-noise generators, namely Rectified Noise (RN), which improves the generative performance by injecting pi-noise into the velocity field of pre-trained RF models. After introducing the Rectified Noise pipeline, pre-trained RF models can be efficiently transformed into pi-noise generators. We validate Rectified Noise by conducting extensive experiments across various model architectures on different datasets. Notably, we find that: (1) RF models using Rectified Noise reduce FID from 10.16 to 9.05 on ImageNet-1k. (2) The models of pi-noise generators achieve improved performance with only 0.39% additional training parameters.