2023-08-01

Model Updating

Incremental Learning = Continual Learning = Model Updating

Paper 1

(zhangNovelMethodDigital2023, link, DOI, zolib)

Anomaly detection algorithm : Isolation Forest

Paper 2 (read this, SOTA on incremental learning)

Three types of incremental learning | Nature Machine Intelligence GMvandeVen/continual-learning: PyTorch implementation of various methods for continual learning (XdG, EWC, SI, LwF, FROMP, DGR, BI-R, ER, A-GEM, iCaRL, Generative Classifier) in three different scenarios. (vandevenThreeTypesIncremental2022, link, DOI, zolib) NeurIPS 2022 Tutorial

Definition • Traditional ML: all training data available at same time • Continual learning:

• training data arrives incrementally
• there is non-stationarity

continual-learning-online-learning

Paper 3 (read this)

Adaptive reconstruction of digital twins for machining systems: A transfer learning approach - ScienceDirect (liuAdaptiveReconstructionDigital2022, link, DOI, zolib)

• Transfer learning
• Vibration

The specific reconstruction steps of the digital twin model are as follows. When the working condition changes, the current working condition changes from the original working condition to the target working condition. A small amount of machining data under the target working condition is collected and preprocessed to obtain the machining data set of the target working condition. Based on the machining data set of the target working condition, the experimental parameters of the mechanism model are finetuned to complete the self-renewal. Then, the primary features of the original data are extracted to obtain the feature data of the target working condition and get stored into the feature database. After this, the distribution distance of feature data between the current working condition and other working condition is calculated to obtain the feature data set with the smallest distribution distance. The data set is indexed in the algorithm model base and machining database to obtain the original model in the algorithm library and the machining data set of related working conditions.

Incremental Learning Library

Creme

• CREME: A toolchain of automatic dataset collection for machine learning in intrusion detection - ScienceDirect
• Incremental machine learning with Creme
• Online/Incremental Learning with Keras and Creme - PyImageSearch

RiverML

• GitHub - online-ml/river: 🌊 Online machine learning in Python

Tornado

• GitHub - alipsgh/tornado: The Tornado framework, designed and implemented for adaptive online learning and data stream mining in Python.

Blogs

• Welcome, AI Astronaut! - Vincenzo Lomonaco

Youtube

• Comparing Incremental Learning Strategies for Convolutional Neural Networks - YouTube
• Few-Shot Class-Incremental Learning - YouTube
• Multi-Domain Incremental Learning for Semantic Segmentation - YouTube
• Yaoyao Liu - Learning from Imperfect Data: Incremental Learning and Few-shot Learning - YouTube

Github

• CristianMorasso/InceptionTimeXICaRL: Time Series Classification with Continual learning using InceptionTime net and ICaRL
• XinShenNu/AI_DeepBL: Codes and data for "An Active Learning-Based Incremental Deep-Broad Learning Algorithm for Unbalanced Time Series Prediction"
• nvcuong/variational-continual-learning: Implementation of the variational continual learning method
• prprbr/awesome-lifelong-continual-learning: A list of papers, blogs, datasets and software in the field of lifelong/continual machine learning
• ContinualAI/continual-learning-papers: Continual Learning papers list, curated by ContinualAI
• optimass/continual_learning_papers: Relevant papers in Continual Learning
• GT-RIPL/Continual-Learning-Benchmark: Evaluate three types of task shifting with popular continual learning algorithms.
• zhoudw-zdw/Awesome-Few-Shot-Class-Incremental-Learning: Awesome Few-Shot Class-Incremental Learning
• xialeiliu/Awesome-Incremental-Learning: Awesome Incremental Learning
• rosenfeldamir/incremental_learning: Initial Code for the paper "incremental learning through deep adaptation"
• shivamsaboo17/Overcoming-Catastrophic-forgetting-in-Neural-Networks: Elastic weight consolidation technique for incremental learning.
• EdenBelouadah/class-incremental-learning
• G-U-N/PyCIL: PyCIL: A Python Toolbox for Class-Incremental Learning
• Vision-Intelligence-and-Robots-Group/Best-Incremental-Learning: An Incremental Learning, Continual Learning, and Life-Long Learning Repository

Arxiv

• 2204.04795.pdf
• (PDF) An Approach for the Incremental Update of a Digital Twin of a Process Plant
• [1612.00796] Overcoming catastrophic forgetting in neural networks
• Task Incremental Learning-Driven Digital-Twin Predictive Modeling for Customized Metal Forming Product Manufacturing Process by Jie Li, Zili Wang, Shuyou Zhang, Yaochen Lin, Jianrong Tan, Lanfang Jiang :: SSRN