In this chapter, we will divide the works into two sections: scientific applications and industrial applications.
(Note that the works list in this pages are works I browsed in background reading. There might be more important works in the corresponding fields not listed here. Besides, I can't ensure the works are representative in the fields. So if you have any comments and recommendations, pls let me know.)
- AL Applications
- Scientific Applications (alphabetical order)
- Industrial Applications (alphabetical order)
- Answer Selection
- Autonomous Driving
- Brain Mapping in a High Performance Computing Environment
- Communication
- Crowd Counting
- Dataset Engineering
- Dialog Policy Learning (intelligent system)
- Disguised Faces Recognition
- Drug Discovery
- Fault Diagnosis of Machinery
- Gas Reservoir Prediction
- Labeling System
- Malware Detection
- Medical Research
- Mobile Health Monitoring / Disease detection
- Person Re-identification
- Privacy Policy Classification
- Questionnaire
- Remote Sensing
- Semiconductor Manufacturing
- Simulation
- Software Engineering
- Solvability Prediction in Power Systems
- Social Bots Detection
- Spam Detection
The scientific applications of AL are about biology, chemistry, physics or things about experiment design and analysis.
Spectroscopic Surveys
- Active deep learning method for the discovery of objects of interest in large spectroscopic surveys: We apply active learning classification methods supported by deep convolutional neural networks to automatically identify complex emission-line shapes in multi-million spectra archives.
- Active learning with RESSPECT: Resource allocation for extragalactic astronomical transients [2020]: The Recommendation System for Spectroscopic follow- up (RESSPECT) project aims to enable the construction of optimized training samples for the Rubin Observatory Legacy Survey of Space and Time (LSST), taking into account a realistic description of the astronomical data environment.
Reduce Simulation Costs
- Active Learning for Computationally Efficient Distribution of Binary Evolution Simulations [2022]
Structural Biology:
- Active machine learning for transmembrane helix prediction [2010, BMC Bioinform]
- Investigating Active Learning and Meta-Learning for Iterative Peptide Design [2020, JCM]
- Large scale active-learning-guided exploration for in vitro protein production optimization [2020, Nature Communications]
- Active learning to classify macromolecular structures in situ for less supervision in cryo-electron tomography [Bioinformatics]
Cell Classification:
- Active feature selection discovers minimal gene-sets for classifying cell-types and disease states in single-cell mRNA-seq data [2021]
Molecular Property Prediction：
- ASGN: An Active Semi-supervised Graph Neural Network for Molecular Property Prediction [2020, KDD]
- Tyger: Task-Type-Generic Active Learning for Molecular Property Prediction [2022]
Chemical natural language processing
- Active Learning Yields Better Training Data for Scientific Named Entity Recognition [2020]
- Active machine learning-driven experimentation to determine compound effects on protein patterns [2016, Elife]
- Categorical Matrix Completion with Active Learning for High-throughput Screening [2020, IEEE Transactions on Computational Biology and Bioinformatics]: Categorical matrix completion is designed to accurately impute the missing experiments while margin sampling is also implemented for uncertainty estimation.
- Active Learning Approach to Optimization of Experimental Control [2020, CHIN. PHYS. LETT.]: Also an optimization problem, where they try to get the optimal control parameters.
- AI-Assisted Scientific Data Collection with Iterative Human Feedback [2021]
- Active Learning for the Optimal Design of Multinomial Classification in Physics [2021]
Lithology Identification:
- Active Domain Adaptation With Application to Intelligent Logging Lithology Identification [IEEE TCYB]
- Evaluation of active learning algorithms for formation lithology identification [2021, Journal of Petroleum Science and Engineering]
Materials Design and Discovery
- Bias free multi-objective active learning for materials design and discovery [2020]
- Active learning for the power factor prediction in diamond-like thermoelectric materials [2020, npj Computational Materials]
- Active Discovery of Catalysts for Sustainable Energy Storage [2021]
- Active discovery of organic semiconductors [2021, Nature communications]
Model checking:
Molecular Dynamics:
- [Active learning accelerates ab initio molecular dynamics on reactive energy surfaces [2020, Chem]
- Active Learning in Bayesian Neural Networks for Bandgap Predictions of Novel Van der Waals Heterostructures [2021, Advanced Intelligent Systems]
- Nanohardness from First Principles with Active Learning on Atomic Environments [2022, JCTC]
- Batch active learning for accelerating the development of interatomic potentials [2022, Comput. Mater. Sci.]
Transition State Calculation:
- Active Learning for Transition State Calculation
- Active Learning of Quantum System Hamiltonians yields Query Advantage [2021]
The industrial applications of AL are about the practical problems or specific requirements in the industry.
Survey Paper:
- Knowledge Augmented Machine Learning with Applications in Autonomous Driving: A Survey [2022]
Line detection:
- Active Learning for Lane Detection: A Knowledge Distillation Approach [2021, ICCV]: Evaluate the uncertainty by the teacher-student models.
Trajectory annotation:
- Engineering Applications of Artificial Intelligence [2022, Engineering Applications of Artificial Intelligence]
About neuron segmentation and tracing at scale.
- Active Learning Pipeline for Brain Mapping in a High Performance Computing Environment [2020, Arxiv]
Eg. 5g, Traffic Classification
- Active Popularity Learning with Cache Hit Ratio Guarantees using a Matrix Completion Committee [2020]: The object is modeling the optimal cache strategy in 5g wireless communication. AL is to learn the content popularities (criterion for determining whether cached or not) since it allows the system to leverage the trade-off between exploration (caching new files) and exploitation (use known files to cache). Specifically, AL is used to do matrix completion on the missing entries of the demand matrix.
- Active content popularity learning and caching optimization with hit ratio guarantees [2020, IEEE Access]: Similar to the last one but with more details. A query is defined as the response received from the user terminal to the system.
- Active Learning for Network Traffic Classification: A Technical Survey [2021, Arxiv]
- Active Sensing for Communications by Learning [2021]
- Active Crowd Counting with Limited Supervision [2020, Arxiv]: An multi-domain crowd counting framework. Use discriminator to learn a feature extractor for the crowd density regression for all the domains.
- Uncertainty Estimation and Sample Selection for Crowd Counting [2020, Arxiv]
- Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop [2015, Arxiv]
- A Simple Yet Brisk And Efficient Active Learning Platform For Text Classification
- Paladin: an annotation tool based on active and proactive learning [2021]
- Scale AI
- Appen (Figure-Eight)
- ClickWorker
- ActiveDeeper: A Model-based Active Data Enrichment System: Use keywords provide by the oracles to search in a database to build a dataset for the current task.
- Models in the Loop: Aiding Crowdworkers with Generative Annotation Assistants [2021]
- Active-learning strategies in computer-assisted drug discovery. [2015, Drug Discov.]
- Active learning for computational chemogenomics [2017, Future Medicinal Chemistry]
- Evaluation of Categorical Matrix Completion Algorithms: Towards Improved Active Learning for Drug Discovery [2021, System Biology]
- Active Learning for Drug Design: A Case Study on the Plasma Exposure of Orally Administered Drugs [2021, J MED CHEM]
- Fault Diagnosis of Rotating Machinery With Limited Expert Interaction: A Multicriteria Active Learning Approach Based on Broad Learning System [2022, ]
This is the most direct industrial application.
Firms:
Papers:
- Q-learning and LSTM based deep active learning strategy for malware defense in industrial IoT applications [2021, Multimedia Tools and Applications]
- Fine-tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally [CVPR, 2017]: The CNN is fine-tuned in each active learning iteration incrementally. Assume that each candidate (predefined) takes one of possible labels Y. This assumption might make it more difficult to generalize.
- Semi-Supervised Active Learning for COVID-19 Lung Ultrasound Multi-symptom Classification [2020]: Multi-label classification network (MSML) is proposed, and a human-machine interaction is exploited to confirm the final annotations that are used to fine-tune MSML with progressively labeled data. Still pool based.
- An Adaptive Low-Rank Modeling-Based Active Learning Method for Medical Image Annotation [2020, IRBM]: Medical images: the intrinsic presence of noise in medical images, the large number of images, and the variety of imaging modalities. Low-rank modeling-based multi-label active learning (LRMMAL) method.
- A Transfer Learning Based Active Learning Framework for Brain Tumor Classification [2020]
- An Active Learning Method for Diabetic Retinopathy Classification with Uncertainty Quantification [2020]
- Deep Reinforcement Active Learning for Medical Image Classification [2020]
- MedSelect: Selective Labeling for Medical Image Classification Combining Meta-Learning with Deep Reinforcement Learning [2021]
- Representative Region Based Active Learning For Histological Classification Of Colorectal Cancer [2021, ISBI]
- Su-Sampling Based Active Learning For Large-Scale Histopathology Image [2021, ICIP]
- PathAL: An Active Learning Framework for Histopathology Image Analysis [2021, T-MI]
- DECAL: DEployable Clinical Active Learning [2022, ICML Workshop]
- Patient Aware Active Learning for Fine-Grained OCT Classification [2022, ICIP]
- Information Gain Sampling for Active Learning in Medical Image Classification [2022]
- Graph Node Based Interpretability Guided Sample Selection for Active Learning [2022, TMI]
- Suggestive annotation: A deep active learning framework for biomedical image segmentation [2017, MICCAI]
- Diminishing Uncertainty within the Training Pool: Active Learning for Medical Image Segmentation [2020, IEEE TRANSACTIONS ON MEDICAL IMAGING]
- Efficient Active Learning for Image Classification and Segmentation Using a Sample Selection and Conditional Generative Adversarial Network [2018, MICCAI]
- Deep Active Learning for Axon-Myelin Segmentation on Histology Data [Arxiv, 2019]
- Labeling Cost Sensitive Batch Active Learning For Brain Tumor Segmentation [2021, ISBI]
- U-Net-Based Active Learning Framework for Enhancing Cancer Immunotherapy [2021, Master's Thesis]
- Active learning for segmentation based on Bayesian sample queries [2021, KBS]
- Reducing Annotating Load: Active Learning with Synthetic Images in Surgical Instrument Segmentation
- Medical symptom recognition from patient text: An active learning approach for long-tailed multilabel distributions [2020, ML4H]
- Deep learning approach for active classification of electrocardiogram signals [2016, Information Science]: (280 citations)
- Human-in-the-loop for efficient training of retinal image analysis methods [2021, Master's Thesis]
- BioAct: Biomedical Knowledge Base Construction using Active Learning [2022]
- Explainable Deep Attention Active Learning for Sentimental Analytics of Mental Disorder [2022, TALLIP]
- Collaborative Multi-Expert Active Learning for Mobile Health Monitoring: Architecture, Algorithms, and Evaluation [2020, Sensors]
- Exploiting Active Learning in Novel Refractive Error Detection with Smartphones [2020, ACMMM]: Use their own created dataset (only 172 images).
- Rethinking data collection for person re-identification: active redundancy reduction [2021, Pattern Recognition]
- Deep Batch Active Learning and Knowledge Distillation for Person Re-identification [2022, IEEE Sensors Journal]
- Deep Active Learning with Crowdsourcing Data for Privacy Policy Classification [2020, Arxiv]
- Vexation-Aware Active Learning for On-Menu Restaurant Dish Availability [2022, KDD]
Remote sensing is an field with really high labeling cost. AL is well used in this field.
Several surveys of AL for remote sensing:
- A survey of active learning algorithms for supervised remote sensing image classification
- A Survey of Active Learning for Quantifying Vegetation Traits from Terrestrial Earth Observation Data [2021]
Papers:
- An Active Deep Learning Approach for Minimally Supervised PolSAR Image Classification [IEEE Transactions on Geoscience and Remote Sensing, 2019]4
- Deep Active Learning for Remote Sensing Object Detection [2020, Arxiv]
- Deep Active Learning in Remote Sensing for data efficient Change Detection [2020, Arxiv]
- Classification of Summer Crops Using Active Learning Techniques on Landsat Images in the Northwest of the Province of Buenos Aires [2020]
- Online Semisupervised Active Classification for Multiview PolSAR Data [2020, IEEE TRANSACTIONS ON CYBERNETICS]
- Hyperspectral Image Classification with Feature-Oriented Adversarial Active Learning [2020, Remote Sensing]
- Adversarial Discriminative Active Deep Learning for Domain Adaptation in Hyperspectral Images Classification [2021, INTERNATIONAL JOURNAL OF REMOTE SENSING]
- Active Deep Learning for Hyperspectral Image Classification With Uncertainty Learning [2021,Active Deep Learning for Hyperspectral Image Classification With Uncertainty Learning]
- Active Learning for Improved Semi-Supervised Semantic Segmentation in Satellite Images [2022, WACV]
- DIAL: Deep Interactive and Active Learning for Semantic Segmentation in Remote Sensing [2021]
- Domain-adaptive active learning for cost-effective virtual metrology modeling [2021, Computers in Industry]
Air Traffic Management (ATM) modeling:
- Active Learning Metamodels for ATM Simulation Modeling [2021, SESAR ]
API Misuse Detection
- A novel framework for detecting social bots with deep neural networks and active learning [2021, KBS]