A new sampling method based on the multivariate Bernoulli distribution has been proposed for multi-label datasets where labels vary significantly in frequency. This novel algorithm accounts for label dependencies by estimating distribution parameters from observed label frequencies and calculating weights for each label combination. Applied to a sample of research articles labeled with 64 biomedical topic categories, the method successfully created a more balanced sub-sample, improving the representation of less common categories while preserving the overall frequency order. AI
RANK_REASON Academic paper detailing a new statistical method for data sampling. [lever_c_demoted from research: ic=1 ai=0.4]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
