Clustered Randomized Smoothing for Stochastic Prediction Functions
TL;DR
Clustered $\alpha$-smoothing is proposed, a framework that partitions noisy samples using an arbitrary clustering algorithm, applies $\alpha$-smoothing locally within each cluster, and combines the resulting predictions into a mixture distribution, derived from interpreting the smoothing distribution as a mixture of $\alpha$-smoothers.
Abstract
Modern stochastic predictors can model rich, multi-modal outcome distributions. However, this expressive power comes with challenges in ensuring robust predictions $-$ a critical requirement in safety-critical domains. Randomized smoothing is a leading technique for improving robustness, particularly against adversarial perturbations. Yet, in stochastic multi-modal regression settings, randomized smoothing often fails due to mode collapse, yielding averaged predictions that do not reflect the underlying distribution. To address this limitation, we propose clustered $\alpha$-smoothing, a framework that (1) partitions noisy samples using an arbitrary clustering algorithm, (2) applies $\alpha$-smoothing locally within each cluster, and (3) combines the resulting predictions into a mixture distribution. By interpreting the smoothing distribution as a mixture of $\alpha$-smoothers, we derive a lower bound on the probability that the smoothed prediction lies within a union of compact regions corresponding to distinct modes. We empirically evaluate our framework on two benchmarks, demonstrating substantial improvements over state-of-the-art methods. In stochastic trajectory prediction on a driving simulator dataset, our approach achieves, on average, a $27\%$ lower Wasserstein distance to the ground-truth distribution compared to $\alpha$-smoothing. In quadrotor control, where modes correspond to distinct feasible paths to a target, our method reduces the collision rate by $81\%$ relative to the state-of-the-art randomized smoothing.