Learning Pipeline

Learning pipelines are multi-stage processes designed to optimize the training of machine learning models, particularly when labeled data is scarce or expensive. Current research focuses on improving pipeline efficiency and performance through techniques like active learning (strategically selecting data for labeling), incorporating diverse data sources (e.g., combining visual and textual information), and developing task-aware strategies that adapt to specific needs (e.g., speech recognition, image analysis). These advancements are significant because they reduce the cost and time associated with data annotation, leading to more efficient and effective model development across various domains, including healthcare, natural language processing, and computer vision.

Papers

June 25, 2024

ALPBench: A Benchmark for Active Learning Pipelines on Tabular Data
Valentin Margraf, Marcel Wever, Sandra Gilhuber, Gabriel Marques Tavares, Thomas Seidl, Eyke Hüllermeier
New Benchmark Active Learning Tabular Data Query Strategy Learning Pipeline

June 14, 2024

GradeADreamer: Enhanced Text-to-3D Generation Using Gaussian Splatting and Multi-View Diffusion
Trapoom Ukarapol, Kevin Pruvost
Gaussian Splatting Diffusion Explainer Text to 3D Generation Multi View Diffusion Digital Asset Partial Credit Grading Learning Pipeline

May 3, 2024

Combining X-Vectors and Bayesian Batch Active Learning: Two-Stage Active Learning Pipeline for Speech Recognition
Ognjen Kundacina, Vladimir Vincan, Dragisa Miskovic
Automatic Speech Recognition Active Learning Speech Recognition Speech Data X Vector Batch Active Learning Learning Pipeline

April 22, 2024

Describe-then-Reason: Improving Multimodal Mathematical Reasoning through Visual Comprehension Training
Mengzhao Jia, Zhihan Zhang, Wenhao Yu, Fangkai Jiao, Meng Jiang
Mathematical Reasoning Visual Understanding Multimodal Mathematical Learning Pipeline

August 5, 2023

Auditing and Robustifying COVID-19 Misinformation Datasets via Anticontent Sampling
Clay H. Yoo, Ashiqur R. KhudaBukhsh
Covid 19 Negative Sampling Audit Evidence Learning Pipeline

November 15, 2022

An Efficient Active Learning Pipeline for Legal Text Classification
Sepideh Mamooler, Rémi Lebret, Stéphane Massonnet, Karl Aberer
Knowledge Distillation Pre Trained Language Model Active Learning Text Classification Fully Supervised Annotation Cost Learning Pipeline

June 30, 2022

End-to-end Learning for Image-based Detection of Molecular Alterations in Digital Pathology
Marvin Teichmann, Andre Aichert, Hanibal Bohnenberger, Philipp Ströbel, Tobias Heimann
Data Detection Digital Pathology Breast Cancer Whole Slide Image Cancer Genome Atlas Learning Pipeline

April 11, 2022

Learning Downstream Task by Selectively Capturing Complementary Knowledge from Multiple Self-supervisedly Learning Pretexts
Jiayu Yao, Qingyuan Wu, Quan Feng, Songcan Chen
Self Supervised Downstream Task Discriminative Representation Self Supervised Pretext Task Learning Pipeline Complementary Representation

April 7, 2022

Task-Aware Active Learning for Endoscopic Image Analysis
Shrawan Kumar Thapa, Pranav Poudel, Binod Bhattarai, Danail Stoyanov
Semantic Segmentation Endoscopic Image Medical Annotation Learning Pipeline

Learning Pipeline

Papers

ALPBench: A Benchmark for Active Learning Pipelines on Tabular Data

GradeADreamer: Enhanced Text-to-3D Generation Using Gaussian Splatting and Multi-View Diffusion

Combining X-Vectors and Bayesian Batch Active Learning: Two-Stage Active Learning Pipeline for Speech Recognition

Describe-then-Reason: Improving Multimodal Mathematical Reasoning through Visual Comprehension Training

Auditing and Robustifying COVID-19 Misinformation Datasets via Anticontent Sampling

An Efficient Active Learning Pipeline for Legal Text Classification

End-to-end Learning for Image-based Detection of Molecular Alterations in Digital Pathology

Learning Downstream Task by Selectively Capturing Complementary Knowledge from Multiple Self-supervisedly Learning Pretexts

Task-Aware Active Learning for Endoscopic Image Analysis