Exploring Invariance in Images through One-way Wave Equations
Yinpeng Chen, Dongdong Chen, Xiyang Dai, Mengchen Liu, Yinan Feng, Youzuo Lin, Lu Yuan, Zicheng Liu
摘要
In this paper, we empirically demonstrate that natural images can be reconstructed with high fidelity from compressed representations using a simple first-order norm-plus-linear autoregressive (FINOLA) process-without relying on explicit positional information. Through systematic analysis, we observe that the learned coefficient matrices (A and B) in FINOLA are typically invertible, and their product, AB -1 , is diagonalizable across training runs. This structure enables a striking interpretation: FINOLA's latent dynamics resemble a system of one-way wave equations evolving in a compressed latent space. Under this framework, each image corresponds to a unique solution of these equations. This offers a new perspective on image invariance, suggesting that the underlying structure of images may be governed by simple, invariant dynamic laws. Our findings shed light on a novel avenue for understanding and modeling visual data through the lens of latent-space dynamics and wave propagation.
问 Lune
几分钟读透这篇 Paper,不用再花一下午。
Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
脉络
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale 引用了 A Simple Framework for Contrastive Learning of Visual Representations
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale 引用了 Language Models are Few-Shot Learners
- Bootstrap Your Own Latent - A New Approach to Self-Supervised Learning 引用了 A Simple Framework for Contrastive Learning of Visual Representations
它引用的 Paper28
- Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah 等NeurIPS 2020 · 64,255 次引用
- A Simple Framework for Contrastive Learning of Visual RepresentationsTing Chen, Simon Kornblith, Mohammad Norouzi, Geoffrey E. HintonICML 2020 · 24,064 次引用
- An Image is Worth 16x16 Words: Transformers for Image Recognition at ScaleAlexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn 等ICLR 2021 · 21,477 次引用
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser 等CVPR 2022 · 13,123 次引用
- Bootstrap Your Own Latent - A New Approach to Self-Supervised LearningJean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec 等NeurIPS 2020 · 9,171 次引用
相关 Paper
- Learning Physics Constrained Dynamics Using AutoencodersTsung-Yen Yang, Justinian Rosca, Karthik Narasimhan, Peter J. RamadgeNeurIPS 2022 · 39 次引用
- High-dimensional neuronal activity from low-dimensional latent dynamics: a solvable modelValentin Schmutz, Ali Haydaroglu, Shuqi Wang, Yixiao Feng 等NeurIPS 2025 · 9 次引用
- Idempotent Learned Image Compression with Right-InverseYanghao Li, Tongda Xu, Yan Wang, Jingjing Liu 等NeurIPS 2023 · 8 次引用
- Fitted Neural Lossless Image CompressionZhe Zhang, Zhenzhong Chen, Shan LiuCVPR 2025
- FARMER: Flow AutoRegressive Transformer over PixelsGuangting Zheng, Qinyu Zhao, Tao Yang, Fei Xiao 等CVPR 2026 · 17 次引用
