Accelerating Learned Image Compression Through Modeling Neural Training Dynamics

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Yichi, Duan, Zhihao, Huang, Yuning, Zhu, Fengqing
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912390095831040
author Zhang, Yichi
Duan, Zhihao
Huang, Yuning
Zhu, Fengqing
author_facet Zhang, Yichi
Duan, Zhihao
Huang, Yuning
Zhu, Fengqing
contents As learned image compression (LIC) methods become increasingly computationally demanding, enhancing their training efficiency is crucial. This paper takes a step forward in accelerating the training of LIC methods by modeling the neural training dynamics. We first propose a Sensitivity-aware True and Dummy Embedding Training mechanism (STDET) that clusters LIC model parameters into few separate modes where parameters are expressed as affine transformations of reference parameters within the same mode. By further utilizing the stable intra-mode correlations throughout training and parameter sensitivities, we gradually embed non-reference parameters, reducing the number of trainable parameters. Additionally, we incorporate a Sampling-then-Moving Average (SMA) technique, interpolating sampled weights from stochastic gradient descent (SGD) training to obtain the moving average weights, ensuring smooth temporal behavior and minimizing training state variances. Overall, our method significantly reduces training space dimensions and the number of trainable parameters without sacrificing model performance, thus accelerating model convergence. We also provide a theoretical analysis on the Noisy quadratic model, showing that the proposed method achieves a lower training variance than standard SGD. Our approach offers valuable insights for further developing efficient training methods for LICs.
format Preprint
id arxiv_https___arxiv_org_abs_2505_18107
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Accelerating Learned Image Compression Through Modeling Neural Training Dynamics
Zhang, Yichi
Duan, Zhihao
Huang, Yuning
Zhu, Fengqing
Image and Video Processing
Computer Vision and Pattern Recognition
As learned image compression (LIC) methods become increasingly computationally demanding, enhancing their training efficiency is crucial. This paper takes a step forward in accelerating the training of LIC methods by modeling the neural training dynamics. We first propose a Sensitivity-aware True and Dummy Embedding Training mechanism (STDET) that clusters LIC model parameters into few separate modes where parameters are expressed as affine transformations of reference parameters within the same mode. By further utilizing the stable intra-mode correlations throughout training and parameter sensitivities, we gradually embed non-reference parameters, reducing the number of trainable parameters. Additionally, we incorporate a Sampling-then-Moving Average (SMA) technique, interpolating sampled weights from stochastic gradient descent (SGD) training to obtain the moving average weights, ensuring smooth temporal behavior and minimizing training state variances. Overall, our method significantly reduces training space dimensions and the number of trainable parameters without sacrificing model performance, thus accelerating model convergence. We also provide a theoretical analysis on the Noisy quadratic model, showing that the proposed method achieves a lower training variance than standard SGD. Our approach offers valuable insights for further developing efficient training methods for LICs.
title Accelerating Learned Image Compression Through Modeling Neural Training Dynamics
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.18107