对抗性细化何时有帮助？将R3GAN适配到时间序列插补的负结果与开放问题
When Does Adversarial Refinement Help? A Negative Result and Open Problem in Adapting R3GAN to Time Series Imputation
- The University of Hong Kong（香港大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本研究将R3GAN适配到时间序列插补，发现对抗性细化在多数配置下无效或有害，并指出判别器无法提供有效细化梯度的原因是一个开放问题。
AI中文摘要：
扩散模型和Transformer在多变量时间序列插补方面已取代GAN，主要原因是GAN训练不稳定。R3GAN（NeurIPS 2024）通过正则化相对损失和可证明的收敛性消除了这种不稳定性，这引发了一个自然的问题：稳定、现代的GAN能否重振对抗性插补？我们将R3GAN适配到1D时间数据，采用从粗到细的细化框架和频域判别器，并在3个数据集上审计了14个保存的配置。由于这些是异构的单次运行，证据是描述性的而非匹配的因果消融。我们报告了一个负结果。所有五个保存的均值/零起始配置改善了48.4%-70.2%。在八个符合条件的非遗留线性起始配置中，平均变化为-0.7%（范围-3.0%至+1.1%）；一个单独的-21.9%遗留日志异常被保留用于溯源，但被排除在该聚合之外。在保存的Weather比较中，独立的R3GAN-1D性能比BRITS差5.8倍。关键的是，我们认为常见的解释（即GAN优化分布目标而非逐点目标）不可能是全部原因，因为扩散模型也优化分布目标却实现了最先进的插补。我们保存的重建权重扫描与对抗信号惰性或有害一致，但无法确定其因果贡献；匹配的去除判别器的消融是关键的下一步实验。我们将学习到的判别器无法提供有用的细化梯度（而学习到的扩散去噪器成功）的确切原因框定为开放问题，并提供关于何时对抗性细化值得的实用指导。
英文摘要：
Diffusion models and transformers have supplanted GANs for multivariate time series imputation, largely on grounds of GAN training instability. R3GAN (NeurIPS 2024) removes that instability via regularized relativistic losses with provable convergence, raising a natural question: do stable, modern GANs revive adversarial imputation? We adapt R3GAN to 1D temporal data with a coarse-to-fine refinement framework and a frequency-domain discriminator, and audit 14 saved configurations across 3 datasets. Because these are heterogeneous single runs, the evidence is descriptive rather than a matched causal ablation. We report a negative result. All five saved mean/zero-start configurations improve by 48.4-70.2%. Among eight eligible non-legacy linear-start configurations, the mean change is -0.7% (range -3.0% to +1.1%); a separate -21.9% legacy logging anomaly is retained for provenance but excluded from that aggregate. In a saved Weather comparison, standalone R3GAN-1D underperforms BRITS by 5.8x. Crucially, we argue the common explanation (that GANs optimize distributional rather than point-wise objectives) cannot be the whole story, since diffusion models also optimize distributional objectives yet achieve state-of-the-art imputation. Our saved reconstruction-weight sweep is consistent with the adversarial signal being inert or harmful, but cannot identify its causal contribution; a matched discriminator-removed ablation is the key next experiment. We frame the precise reason a learned discriminator fails to provide useful refinement gradients (where a learned diffusion denoiser succeeds) as an open problem, and offer practical guidance on when adversarial refinement is worthwhile.
