This repo contains the paper list and figures for A Survey of Data Drift and Model Adaptation in Industrial Settings.
总体框架~~~
The scope of this survey is mainly defined by following aspects.
- xx
@article{chen2024a,
title = {A Survey of Data Drift and Model Adaptation in Industrial Settings},
author = {Chen, Jiao and Liu, Qianmiao and Dai, Suyan and He, Jiayi
and Lv, Zuohong},
journal={arXiv preprint arXiv:240x.xxxxx},
year = {2024}
}
-
- Industrial Fault Diagnosis
- Remaining Life Prediction
- Laser Micro/Nano Processing
- Planing and Control for Autonomous Driving
- Biomorphic Robotic Motion Control
- Additive Manufacturing Process Monitoring and Control
- Gesture Recognition Based on Hydrogel Electronic Skin
-
[Short-term Drift Adaptation Strategies](#Short-term Drift Adaptation Strategies)
- Before Deployment
- After Deployment
-
[Long-term Drift Adaptation Strategies](#Long-term Drift Adaptation Strategies)
- Continual/Lifelong Learning
- Learn from Model
-
- Framework/Platform Development
- Datasets/Benchmarks
- Integration with Large Models
- Multi-Model Management
- Knowledge Base Construction
(工业故障诊断)
(剩余寿命预测)
(激光微纳加工)
(自动驾驶规划与控制,e.g., 自动清扫车)
(仿生机器人运动控制)
(增材制造过程监控及控制)
(水凝胶电子皮肤应用、e.g., 手势识别)
The goal is to present a simple and intuitive overview of the definition, types, and case studies of data drift. Introducing related concepts: Out-Of-Distribution (OOD), Long-Tail Distribution, Non-IID, and Few-Shot Learning.
TODO: 给出一个示意图,代表不同的检测方法在cifar10数据集的检测效果。
Concept Drift Adaptation by exploiting Drift Type.[ACM Transactions on Knowledge Discovery from Data 2023][[paper](Concept Drift Adaptation by Exploiting Drift Type | ACM Transactions on Knowledge Discovery from Data)]
Flower: A friendly federated learning research framework. [arXiv'20] [Paper] [Code]
Fedml: A research library and benchmark for federated machine learning. [arXiv'20] [Paper] [Code]
[Serving on Edge]
e.g., Prompt Adapter, LoRA, Prefix tuning
e.g., Edge-Cloud Collaboration (by Chen), Test-Time Adaptation, Transfer Learning
EdgeFM: Leveraging Foundation Model for Open-set Learning on the Edge. [SenSys'23] [Paper]
e.g., Data Selection Mechanisms Based on Gradients, Entropy, Data Annotation (Language Models, Pseudo-Label Generation)
[Serving on Cloud]
Towards Edge-Cloud Collaborative Machine Learning: A Quality-aware Task Partition Framework.[paper]
ECLM: Efficient Edge-Cloud Collaborative Learning with Continuous Environment Adaptation.[paper]
Memory efficient continual learning with transformers. [NeurIPS'22] [Paper]