Understanding Untrained Deep Models for Inverse Problems [From the Guest Editors]: Algorithms and theory

计算机科学 反问题 算法 人工智能 数学优化 反向 信号处理 应用数学 数学 人工神经网络 估计理论 趋同(经济学) 理论计算机科学 计算复杂性理论
作者
Ismail Alkhouri,Evan A. Bell,Avrajit Ghosh,Shijun Liang,Rongrong Wang,Saiprasad Ravishankar
出处
期刊:IEEE Signal Processing Magazine [Institute of Electrical and Electronics Engineers]
卷期号:43 (2): 77-92 被引量:15
标识
DOI:10.1109/msp.2025.3632786
摘要

In recent years, deep learning (DL) methods have been extensively developed for inverse imaging problems (IIPs), encompassing supervised, self-supervised, and generative approaches. Most of these methods require large numbers of labeled or unlabeled training data to learn effective models. However, in many practical applications, such as medical image reconstruction, extensive training datasets are often unavailable or limited. A significant milestone in addressing this challenge came in 2018 with the work of Ulyanov et al.[1], which introduced the deep image prior (DIP)—the first training-data-free convolutional neural network (CNN) method for IIPs. Unlike conventional DL approaches, DIP requires only a CNN, the noisy measurements, and a forward operator. By leveraging the implicit regularization of deep networks initialized with random noise, DIP can learn and restore image structures without relying on external datasets. However, a well-known limitation of DIP is its susceptibility to overfitting, primarily due to overparameterization of the network. In this tutorial article, we provide a comprehensive review of DIP, including a theoretical analysis of its training dynamics. We also categorize and discuss recent advancements in DIP-based methods that are aimed at mitigating overfitting, including techniques such as regularization, network reparameterization, and early stopping. Furthermore, we discuss approaches that combine DIP with pretrained neural networks (NNs), present empirical comparison results against data-centric methods, and highlight open research questions and future directions.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
荼蘼发布了新的文献求助10
1秒前
沉静忘幽发布了新的文献求助10
1秒前
852的应助被天穹雨采纳,获得10
1秒前
Orange的应助被老年学术废物采纳,获得10
2秒前
4秒前
舒适忆曼完成签到,获得积分10
4秒前
4秒前
lZzz发布了新的文献求助10
5秒前
5秒前
孤独盼望完成签到,获得积分10
5秒前
7秒前
soufle完成签到,获得积分10
8秒前
8秒前
DW的应助被捏鱼采纳,获得10
9秒前
搞怪的芾完成签到,获得积分10
9秒前
9秒前
Richardisme完成签到,获得积分10
9秒前
LIU发布了新的文献求助10
10秒前
荼蘼发布了新的文献求助10
10秒前
11秒前
11秒前
共享精神的应助被自觉的以寒采纳,获得10
12秒前
ppsy完成签到,获得积分10
12秒前
lll发布了新的文献求助10
12秒前
12秒前
tutu发布了新的文献求助10
12秒前
ket的应助被chen采纳,获得10
12秒前
甜甜的满天完成签到,获得积分10
13秒前
小二郎的应助被W1ndyKnight采纳,获得10
13秒前
13秒前
思源的应助被AA采纳,获得10
14秒前
XXX完成签到 ,获得积分10
15秒前
萍萍完成签到 ,获得积分10
15秒前
溺水小刀发布了新的文献求助10
16秒前
100发布了新的文献求助10
16秒前
nini发布了新的文献求助10
16秒前
syvshc的应助被小燕尾条采纳,获得10
16秒前
Yes完成签到,获得积分10
17秒前
李健的应助被浮晓采纳,获得10
17秒前
Eli发布了新的文献求助10
18秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
John McDowells Bildungstheorie 800
八旗制度の研究 600
plantTFDB 508
Resource Recovery Handbook 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8175157
求助须知:如何正确求助?哪些是违规求助? 9566975
关于积分的说明 21789458
捐赠科研通 7769647
什么是DOI,文献DOI怎么找? 3431836
关于科研通互助平台的介绍 2556078
邀请新用户注册赠送积分活动 2477241