New ML Framework for Learning with Inaccurate True Targets

A novel machine learning framework named EL-MIATTs (Evaluation and Learning with Multiple Inaccurate True Targets) has been introduced to address situations where the true target is ambiguous or subjective. It operates under the premise that an objective true target may not exist. This research outlines two complementary approaches: evaluation algorithms based on the Logical Assessment Formula (LAF) and learning strategies derived from Undefinable True Target Learning (UTTL) utilizing MIATTs. These methods facilitate coherent and practical modeling in uncertain environments. The authors investigate task-specific MIATTs, focusing on how their diversity and coverage affect structural characteristics and influence subsequent evaluation and learning. This study connects theoretical insights with practical applications of EL-MIATTs.