Awesome

python-glad

A Python Implementation of GLAD.

GLAD

Algorithm for aggregating labels given by labeler more acculately than majority vote.

• Whitehill, Jacob, et al. "Whose vote should count more: Optimal integration of labels from labelers of unknown expertise." Advances in neural information processing systems. 2009.

Dependency

• numpy
• scipy

Usage

$ python glad.py data/data.txt

Input Format

  First line:
    <numGivenLabels> <numLabelers> <numTasks> <Prior p(Z=1)>
  Following <numGivenLabels> lines:
    <taskId> <labelerId> <label:0|1>

• The task IDs must be integers and must be 0...<numImages-1>.
• The labeler IDs must be integers and must be 0...<numLabelers-1>.