Streaming Active Learning with Deep Neural Networks

Akanksha Saran, Safoora Yousefi, Akshay Krishnamurthy, John Langford, Jordan T. Ash
3/5/2023
cs.LG

No official code found. Best implementation: asaran/VeSSAL (14 stars, last commit 2025-03-07).

Abstract

Active learning is perhaps most naturally posed as an online learning problem. However, prior active learning approaches with deep neural networks assume offline access to the entire dataset ahead of time. This paper proposes VeSSAL, a new algorithm for batch active learning with deep neural networks in streaming settings, which samples groups of points to query for labels at the moment they are encountered. Our approach trades off between uncertainty and diversity of queried samples to match a desired query rate without requiring any hand-tuned hyperparameters. Altogether, we expand the applicability of deep neural networks to realistic active learning scenarios, such as applications relevant to HCI and large, fractured datasets.

View on arXivView PDF

Code Implementations(1)

asaran/VeSSALCommunity100%

This repository contains code used to conduct experiments reported in the paper "Streaming Active Learning with Deep Neural Networks" accepted at ICML 2023.

146PythonFeb 19, 20231 years ago

Similar Papers

Deep Active Learning with Contrastive Learning Under Realistic Data Pool Assumptions

83% match2023

Accelerating Batch Active Learning Using Continual Learning Techniques

82% match2023

Combining self-labeling and demand based active learning for non-stationary data streams

80% match2023

Active Learning Guided by Efficient Surrogate Learners

80% match2023

A Survey on Deep Active Learning: Recent Advances and New Frontiers

80% match2024

Neural Active Learning Meets the Partial Monitoring Framework

80% match2024

Cite this paper

@article{saran2023streaming,
  title  = {Streaming Active Learning with Deep Neural Networks},
  author = {Akanksha Saran and Safoora Yousefi and Akshay Krishnamurthy and John Langford and Jordan T. Ash},
  year   = {2023},
  eprint = {2303.02535},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2303.02535v2},
  journal = {ICML 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2303.02535.svg)](https://codeofpaper.com/paper/2303.02535)

Discussion