Technology Category
Computer Vision: Learning & Optimization for CVMachine Learning: Learning on the Edge & Model CompressionSearch and Optimization: Learning to Search
Leveraging Second-Order Curvature for Efficient Learned Image Compression: Theory and Empirical Evidence
🤖 AI Summary
This work addresses the slow convergence and performance limitations in learned image compression, which often arise from gradient conflicts between rate and distortion objectives during training. For the first time, the study introduces the second-order quasi-Newton optimizer SOAP as a plug-and-play module into this domain, effectively mitigating optimization conflicts and substantially improving both training efficiency and stability. Theoretical analysis and extensive experiments demonstrate that SOAP not only accelerates convergence across multiple state-of-the-art compression models but also significantly reduces outliers in activations and latent variables. This leads to enhanced robustness for post-training quantization and consistent improvements in rate–distortion performance.
Technology Category
Application Category
📝 Abstract
Training learned image compression (LIC) models entails navigating a challenging optimization landscape defined by the fundamental trade-off between rate and distortion. Standard first-order optimizers, such as SGD and Adam, struggle with \emph{gradient conflicts} arising from competing objectives, leading to slow convergence and suboptimal rate-distortion performance. In this work, we demonstrate that a simple utilization of a second-order quasi-Newton optimizer, \textbf{SOAP}, dramatically improves both training efficiency and final performance across diverse LICs. Our theoretical and empirical analyses reveal that Newton preconditioning inherently resolves the intra-step and inter-step update conflicts intrinsic to the R-D objective, facilitating faster, more stable convergence. Beyond acceleration, we uncover a critical deployability benefit: second-order trained models exhibit significantly fewer activation and latent outliers. This substantially enhances robustness to post-training quantization. Together, these results establish second-order optimization, achievable as a seamless drop-in replacement of the imported optimizer, as a powerful, practical tool for advancing the efficiency and real-world readiness of LICs.
Problem
Research questions and friction points this paper is trying to address.
learned image compression
rate-distortion trade-off
gradient conflicts
optimization landscape
training convergence
Innovation
Methods, ideas, or system contributions that make the work stand out.
second-order optimization
learned image compression
rate-distortion trade-off
gradient conflict
post-training quantization
🔎 Similar Papers
No similar papers found.
