arXiv：cs.LG· Thejani Gamage, Hyemin Gu, Zhizhen Zhang, Ziyu Chen, Markos A. Katsoulakis, Luc Rey-Bellet·· 6 小时前AI 评分39
CVaR-GPA：用 CVaR 惩罚的 Wasserstein 梯度流微调生成模型以建模极端事件
Fine-Tuning Generative Models for Extreme Events via CVaR-Penalized Wasserstein Gradient Flows
AI 导读
研究者提出 CVaR-GPA，一种通过 CVaR 惩罚的 Wasserstein 梯度流微调生成模型、使其逼近重尾分布的算法，无需访问预训练模型内部架构。在包括 Ohio River 流域日径流（d=64）和 Fama-French 投资组合（d=25）在内的四个目标上，该方法用单组超参数将全局误差和尾部误差分别降低 14.0× 和 9.8×，覆盖 GAN、扩散模型等七种预训练模型。
正文
Abstract:In many high-stakes domains, extreme events carry substantial consequences, yet learning the heavy-tailed distributions that govern them from finite samples remains challenging: the quantities of interest are driven by a few extreme observations, so the tail is under-sampled relative to its importance. Even generative models tailored for heavy tails capture the tail region inadequately in practice. We propose the Conditional Value-at-Risk (CVaR)-penalized Generative Particle Algorithm (CVaR-GPA), a tail-agnostic algorithm for fine-tuning generative models toward heavy-tailed targets, built as a time discretization of the Wasserstein gradient flow of the Lipschitz-regularized KL divergence penalized by a CVaR discrepancy term. Such a flow can be initialized from the output samples of the pre-trained model, which are then transported along the gradient descent of the loss functional, without requiring access to the pre-trained model's internal architecture. The Lipschitz-regularized KL divergence requires minimal assumptions on the target, while the CVaR penalty focuses the flow on the tail. The CVaR penalty depends on the target only through a scalar tail statistic, inducing a velocity field that remains active in the under-sampled tail region at a dimension-free estimation cost. The resulting velocity field, and hence the depth of the transport map, implicitly adapts to the target, without target-specific modifications. Across four targets, including two real-world, high-dimensional datasets (daily streamflow in the Ohio River basin ($d=64$) and the Fama-French portfolios ($d=25$) with tail indices ranging from $1.05$ to $3.34$), fine-tuning with CVaR-GPA reduces global and tail errors by geometric-mean factors of $14.0 \times$ and $9.8 \times$, respectively, across seven pre-trained models spanning GANs, diffusion models, and other generative flows, with a single set of hyperparameters.
| Subjects: | Machine Learning (stat.ML); Machine Learning (cs.LG) |
| Cite as: | arXiv:2608.11544 [stat.ML] |
| (or arXiv:2608.11544v2 [stat.ML] for this version) | |
| https://doi.org/10.48550/arXiv.2608.11544 arXiv-issued DOI via DataCite |
Submission history
From: Thejani Gamage [view email]
[v1]
Wed, 12 Aug 2026 01:26:14 UTC (295 KB)
[v2]
Thu, 1 Oct 2026 22:08:16 UTC (1,386 KB)
来源：arXiv：cs.LG · arxiv.org