Novel Active Learning

Novel active learning methods aim to significantly reduce the cost of labeling data for training machine learning models by strategically selecting the most informative samples for annotation. Current research focuses on addressing the "cold start" problem (poor initial performance with limited labeled data) through techniques like leveraging large language models for sample selection, dynamically adjusting strategies based on model competence, and focusing on data points near decision boundaries or representing minority classes. These advancements are improving model performance across various tasks, including image classification, semantic segmentation, and natural language processing, leading to more efficient and effective machine learning in data-scarce scenarios.

Papers

July 1, 2024

DCoM: Active Learning for All Learners
Inbal Mishal, Daphna Weinshall
LeArning Abstract Active Learning Deep Model Deep Active Learning Margin Based Coverage Control Novel Active Learning

May 17, 2024

ActiveLLM: Large Language Model-based Active Learning for Textual Few-Shot Scenarios
Markus Bayer, Christian Reuter
Active Learning Text Modality Multi Scenario Novel Active Learning

March 6, 2024

SUPClust: Active Learning at the Boundaries
Yuta Ono, Till Aczel, Benjamin Estermann, Roger Wattenhofer
Active Learning Decision Boundary Document Boundary Novel Active Learning Decision Region

November 23, 2023

Class Balanced Dynamic Acquisition for Domain Adaptive Semantic Segmentation using Active Learning
Marc Schachtsiek, Simone Rossi, Thomas Hannagan
Semantic Segmentation Active Learning Domain Adaptive Semantic Segmentation Label Efficient Novel Active Learning Acquisition Model

August 15, 2023

Graph-Structured Kernel Design for Power Flow Learning using Gaussian Processes
Parikshit Pareek, Deepjyoti Deka, Sidhant Misra
Gaussian Process Graph Kernel Power Flow Active Learning Algorithm Novel Active Learning

June 7, 2023

NTKCPL: Active Learning on Top of Self-Supervised Model by Estimating True Coverage
Ziting Wen, Oscar Pizarro, Stefan Williams
Self Supervised Active Learning Estimation Task Neural Tangent Kernel Self Supervised Model Higher Coverage Rate ActIve Learning algORithm Novel Active Learning

January 12, 2023

Forgetful Active Learning with Switch Events: Efficient Sampling for Out-of-Distribution Data
Ryan Benkert, Mohit Prabhushankar, Ghassan AlRegib
Active Learning Anti Forgetting Distribution Data Switch List Efficient Sampling Outlier Removal Novel Active Learning Deep Out of Distribution

December 6, 2022

MUS-CDB: Mixed Uncertainty Sampling with Class Distribution Balancing for Active Annotation in Aerial Object Detection
Dong Liang, Jing-Wei Zhang, Ying-Peng Tang, Sheng-Jun Huang
Active Learning Aerial Object Detection Uncertainty Sampling Novel Active Learning Class Balancing Technique

October 3, 2022

Nonstationary data stream classification with online active learning and siamese neural networks
Kleanthis Malialis, Christos G. Panayiotou, Marios M. Polycarpou
Classification Code Active Learning Siamese Network Novel Active Learning Non Stationary Data

January 19, 2022

PT4AL: Using Self-Supervised Pretext Tasks for Active Learning
John Seon Keun Yi, Minseok Seo, Jongchan Park, Dong-Geol Choi
Active Learning Self Supervised Pretext Task Novel Active Learning

December 5, 2021

Multi-View Active Learning for Short Text Classification in User-Generated Data
Payam Karisani, Negin Karisani, Li Xiong
Active Learning Multi View Learning Twitter Dataset Short Text Novel Active Learning

Novel Active Learning

Papers

DCoM: Active Learning for All Learners

ActiveLLM: Large Language Model-based Active Learning for Textual Few-Shot Scenarios

SUPClust: Active Learning at the Boundaries

Class Balanced Dynamic Acquisition for Domain Adaptive Semantic Segmentation using Active Learning

Graph-Structured Kernel Design for Power Flow Learning using Gaussian Processes

NTKCPL: Active Learning on Top of Self-Supervised Model by Estimating True Coverage

Forgetful Active Learning with Switch Events: Efficient Sampling for Out-of-Distribution Data

MUS-CDB: Mixed Uncertainty Sampling with Class Distribution Balancing for Active Annotation in Aerial Object Detection

Nonstationary data stream classification with online active learning and siamese neural networks

PT4AL: Using Self-Supervised Pretext Tasks for Active Learning

Multi-View Active Learning for Short Text Classification in User-Generated Data