![[1803.09319] SUNLayer: Stable denoising with generative networks](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[1803.09319] SUNLayer: Stable denoising with generative networks
Summary
The paper introduces SUNLayer, a theoretical framework for stable denoising using generative networks, focusing on activation functions and their optimization in deep learning models.
Why It Matters
This research addresses the challenges of image denoising and other inverse problems in machine learning by providing a robust theoretical foundation. By identifying conditions for effective activation functions, it enhances the reliability of generative models, which are crucial in various applications, including computer vision and data recovery.
Key Takeaways
- SUNLayer offers a new theoretical framework for generative models.
- Explicit conditions on activation functions are identified for effective denoising.
- Numerical experiments validate the framework's stability in denoising tasks.
- The research contributes to the understanding of generative networks in deep learning.
- Applications extend to image denoising, compressed sensing, and super-resolution.
Computer Science > Machine Learning arXiv:1803.09319 (cs) [Submitted on 25 Mar 2018 (v1), last revised 20 Feb 2026 (this version, v2)] Title:SUNLayer: Stable denoising with generative networks Authors:Ruhui Jin, Dustin G. Mixon, Soledad Villar View a PDF of the paper titled SUNLayer: Stable denoising with generative networks, by Ruhui Jin and 2 other authors View PDF HTML (experimental) Abstract:Deep neural networks are often used to implement powerful generative models for real-world data. Notable applications include image denoising, as well as other classical inverse problems like compressed sensing and super-resolution. To provide a rigorous but simplified analysis of generative models, in this work, we introduce an elegant theoretical framework based on spherical harmonics, namely \textbf{SUNLayer}. Our theoretical framework identifies explicit conditions on activation functions that guarantee denoising under local optimization. Numerical experiments examine the theoretical properties on commonly used activation functions and demonstrate their stable denoising performance. Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:1803.09319 [cs.LG] (or arXiv:1803.09319v2 [cs.LG] for this version) https://doi.org/10.48550/arXiv.1803.09319 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Ruhui Jin [view email] [v1] Sun, 25 Mar 2018 19:33:04 UTC (2,042 KB) [v2] Fri, 20 Feb 2026 15:59:36 UTC (5,647 KB) Full-text links: Acc...