Summary of Neural Active Learning Meets the Partial Monitoring Framework, by Maxime Heuillet et al.
Neural Active Learning Meets the Partial Monitoring Framework
by Maxime Heuillet, Ola Ahmad, Audrey Durand
First submitted to arxiv on: 14 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed foundation for online-based active learning (OAL) tasks is built upon partial monitoring, a theoretical framework specialized in online learning from partially informative actions. The authors show that previously studied binary and multi-class OAL tasks are instances of partial monitoring. Additionally, they introduce a new class of cost-sensitive OAL tasks. A novel neural network-based strategy called NeuralCBP is proposed to account for predictive uncertainty. Empirical evaluation on open-source datasets demonstrates the favorable performance of NeuralCBP compared to state-of-the-art baselines across multiple binary, multi-class, and cost-sensitive OAL tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary OAL allows an agent to learn from a stream of observations while balancing the cost of getting more information (labeled data) with the cost of making mistakes. The authors create a new way for OAL by using “partial monitoring,” which is good at learning from actions that are only partially informative. They show that some previous OAL tasks fit into this framework, and they also introduce new types of tasks that consider the cost of different kinds of errors. To help with these new tasks, the authors create a neural network called NeuralCBP that takes uncertainty into account. This approach performs well on many real-world datasets. |
Keywords
» Artificial intelligence » Active learning » Neural network » Online learning
