Mapping Networks: CVPR 2026 Best Paper Award Nominee
The paper titled Mapping Networks has been nominated for the CVPR 2026 Best Paper Award. The paper introduces a new concept called Mapping Networks, which aims to reduce overfitting in deep learning models by replacing high-dimensional weight space with a compact, trainable latent vector. The authors, Lord Sen and Shyamapada Mukherjee, demonstrate that Mapping Networks can achieve comparable or better performance than target networks with a significant reduction in trainable parameters.
- ▪The paper Mapping Networks has been nominated for the CVPR 2026 Best Paper Award.
- ▪Mapping Networks replace high-dimensional weight space with a compact, trainable latent vector.
- ▪The approach achieves a 99.5% reduction in trainable parameters and comparable or better performance than target networks.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.19134 (cs) [Submitted on 22 Feb 2026] Title:Mapping Networks Authors:Lord Sen, Shyamapada Mukherjee View a PDF of the paper titled Mapping Networks, by Lord Sen and 1 other authors View PDF HTML (experimental) Abstract:The escalating parameter counts in modern deep learning models pose a fundamental challenge to efficient training and resolution of overfitting. We address this by introducing the \emph{Mapping Networks} which replace the high dimensional weight space by a compact, trainable latent vector based on the hypothesis that the trained parameters of large networks reside on smooth, low-dimensional manifolds.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
