# SaluNet: Can Learned Geometry Replace Conventional Feature Normalization? **URL:** \ **Category:** 🤗Transformers\ **Created:** [November 9, 2026, 5:58pm UTC](https://discuss.huggingface.co/t/salunet-can-learned-geometry-replace-conventional-feature-normalization/191606 "2026-11-09T17:58:05Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![Rehab-92](https://avatars.discourse-cdn.com/v4/letter/r/cab0a1/32.png) [@Rehab-92](https://discuss.huggingface.co/u/Rehab-92)\ **Post date:** [November 9, 2026, 5:58pm UTC](https://discuss.huggingface.co/t/salunet-can-learned-geometry-replace-conventional-feature-normalization/191606/1 "2026-11-09T17:58:05Z")
I would like to highlight an interesting research contribution by **Prof. Mourad Zaied** , introducing SaluNet, a framework that explores an alternative perspective on normalization in modern deep learning architectures. The central idea is particularly interesting: instead of relying on batch-wise or activation statistics, can learnable transformations provide an effective alternative to conventional normalization layers? The proposed framework introduces three key components: - **SALU:** A learnable elementwise transformation designed to replace conventional normalization layers. - **SWALU and GALU:** Learnable activation alternatives inspired by Swish and GELU. - **SaluNet-C and SaluNet-T:** Extensions of this approach to Convolutional Neural Networks and Vision Transformers. The reported experiments on CIFAR-100 and ImageNet-1K explore the framework’s performance under different batch-size configurations. What I find particularly interesting is the broader research direction: rethinking fundamental architectural choices in deep learning rather than simply modifying existing components. **Resources:** 📄 Paper: [LinkedIn](https://lnkd.in/e_m2SWb5) 💻 Code, configurations and pretrained models: [LinkedIn](https://lnkd.in/en7_SSp8) 🐦 Original announcement: I would be interested in hearing the community’s thoughts: Could learnable geometric transformations offer a promising direction for reducing the dependence of deep learning architectures on conventional normalization techniques? #DeepLearning #ComputerVision #🤗Transformers #CNN #Research ---
**Author:** ![discourse\_ai\_spam](https://avatars.discourse-cdn.com/v4/letter/d/c68b51/32.png) [@discourse\_ai\_spam](https://discuss.huggingface.co/u/discourse_ai_spam)\ **Post date:** [November 9, 2026, 5:58pm UTC](https://discuss.huggingface.co/t/salunet-can-learned-geometry-replace-conventional-feature-normalization/191606/2 "2026-11-09T17:58:07Z")
Unlisted on November 9, 2026, 5:58pm UTC ---
**Author:** ![system](https://us1.discourse-cdn.com/hellohellohello/original/2X/d/de4155eb4aa4108ecb32a1389d7cc37ae69f88b7.png) [@system](https://discuss.huggingface.co/u/system)\ **Post date:** [November 10, 2026, 10:56am UTC](https://discuss.huggingface.co/t/salunet-can-learned-geometry-replace-conventional-feature-normalization/191606/3 "2026-11-10T10:56:24Z")
Listed on November 10, 2026, 10:56am UTC