Start Active Learning

Start active learning addresses the challenge of training machine learning models when no labeled data is initially available, a "cold-start" problem common in many applications. Current research focuses on developing effective strategies for selecting initial data subsets for model initialization, often employing clustering techniques enhanced by foundation models or leveraging proxy tasks and size-balanced sampling to mitigate class imbalance. These advancements aim to significantly reduce the annotation effort required for training high-performing models, impacting fields like medical image analysis, natural language processing, and recommendation systems by enabling efficient model development with limited labeled data.

Papers

July 24, 2024

A Novel Two-Step Fine-Tuning Pipeline for Cold-Start Active Learning in Text Classification Tasks
Fabiano Belém, Washington Cunha, Celso França, Claudio Andrade, Leonardo Rocha, Marcos André Gonçalves
Active Learning Long Form Novel Text Classification Task BERT Embeddings Start Active Learning Text Representation Method

February 21, 2024

STENCIL: Submodular Mutual Information Based Weak Supervision for Cold-Start Active Learning
Nathan Beck, Adithya Iyer, Rishabh Iyer
Large Language Model Active Learning Large Corpus Text Classification Submodular Mutual Information Start Active Learning

February 4, 2024

Foundation Model Makes Clustering A Better Initialization For Cold-Start Active Learning
Han Yuan, Chuan Hong
Foundation Model Active Learning Self Supervised Paradigm Better Initialization Start Active Learning

October 12, 2023

BaSAL: Size-Balanced Warm Start Active Learning for LiDAR Semantic Segmentation
Jiarong Wei, Yancong Lin, Holger Caesar
Semantic Segmentation Active Learning Mid Range LiDAR Lidar Semantic Segmentation Human Base Start Active Learning

July 22, 2023

COLosSAL: A Benchmark for Cold-start Active Learning for 3D Medical Image Segmentation
Han Liu, Hao Li, Xing Yao, Yubo Fan, Dewei Hu, Benoit Dawant, Vishwesh Nath, Zhoubing Xu, Ipek Oguz
New Benchmark Annotated Dataset 3D Medical Image Segmentation Medical Segmentation Decathlon Mini Giant Start Active Learning

April 23, 2023

You Never Get a Second Chance To Make a Good First Impression: Seeding Active Learning for 3D Semantic Segmentation
Nermin Samet, Oriane Siméoni, Gilles Puy, Georgy Ponimatkin, Renaud Marlet, Vincent Lepetit
Semantic Segmentation Point Cloud Active Learning 3D Point Cloud 3D Semantic Segmentation Efficient Annotation Impression Generation Start Active Learning

October 1, 2022

Speed Up the Cold-Start Learning in Two-Sided Bandits with Many Arms
Mohsen Bayati, Junyu Cao, Wanning Chen
Multi Armed Bandit Speed Effect Online Experiment Randomized Experiment Multi Arm Bandit Start Active Learning Many Arm

September 15, 2022

Cold-Start Data Selection for Few-shot Language Model Fine-tuning: A Prompt-Based Uncertainty Propagation Approach
Yue Yu, Rongzhi Zhang, Ran Xu, Jieyu Zhang, Jiaming Shen, Chao Zhang
Language Model Pre Trained Language Model Prompt Based Learning Uncertainty Propagation Start Active Learning

September 13, 2022

Warm Start Active Learning with Proxy Labels \& Selection via Semi-Supervised Fine-Tuning
Vishwesh Nath, Dong Yang, Holger R. Roth, Daguang Xu
Semi Supervised Medical Image Datasets Supervised Fine Tuning View Selection Medical Segmentation Decathlon Iteration Head Start Active Learning Proxy Label

January 25, 2022

Cold Start Active Learning Strategies in the Context of Imbalanced Classification
Etienne Brangbour, Pierrick Bruneau, Thomas Tamisier, Stéphane Marchand-Maillet
Active Learning Context Information Class Imbalance Label Propagation Imbalanced Classification Limited Label Start Active Learning

Start Active Learning

Papers

A Novel Two-Step Fine-Tuning Pipeline for Cold-Start Active Learning in Text Classification Tasks

STENCIL: Submodular Mutual Information Based Weak Supervision for Cold-Start Active Learning

Foundation Model Makes Clustering A Better Initialization For Cold-Start Active Learning

BaSAL: Size-Balanced Warm Start Active Learning for LiDAR Semantic Segmentation

COLosSAL: A Benchmark for Cold-start Active Learning for 3D Medical Image Segmentation

You Never Get a Second Chance To Make a Good First Impression: Seeding Active Learning for 3D Semantic Segmentation

Speed Up the Cold-Start Learning in Two-Sided Bandits with Many Arms

Cold-Start Data Selection for Few-shot Language Model Fine-tuning: A Prompt-Based Uncertainty Propagation Approach

Warm Start Active Learning with Proxy Labels \& Selection via Semi-Supervised Fine-Tuning

Cold Start Active Learning Strategies in the Context of Imbalanced Classification