Reproduce “Error Rate Bounds and Iterative Weighted Majority Voting for Crowdsourcing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hongwei Li, Bin Yu
A theoretical framework establishes finite-sample error rate bounds for general crowdsourcing label aggregation rules. Building on this, the Iterative Weighted Majority Voting (IWMV) algorithm is introduced, demonstrating comparable accuracy to existing methods with notably faster computation on real-world datasets.
Key insights
- Error rates decrease exponentially with the number of workers when there is sufficient separation between classes, a relationship quantified by finite-sample bounds.
- An 'oracle bound-optimal rule' for setting worker weights, derived from theoretical analysis, closely approximates the maximum a posteriori (MAP) rule.
- The Iterative Weighted Majority Voting (IWMV) algorithm iteratively refines label estimates and worker weights, offering one-step performance guarantees and computational efficiency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.