Logistic Schedule

The logistic schedule applies the logistic function to shape the noise trajectory. Like cosine and sigmoid schedules, it produces a smooth S-shaped transition, but is parameterized via a steepness coefficient $k$ and a midpoint $t_0$. Noise is introduced slowly at the beginning, followed by a rapid increase around the midpoint, and then smooths out again as $t$ approaches $T$. By preserving the integrity of the original data distribution early in the process, the logistic schedule contributes to improved training stability and overall generative performance. Its smooth continuous dynamics minimize unintended deviations, resulting in more accurate latent predictions.