Human-in-the-Loop Mixup: Aligning Synthetic Data with Human Perception - Oral Session 2

Explore a 26-minute conference talk from the Uncertainty in Artificial Intelligence (UAI) 2023 Oral Session 2 focusing on "Human in the Loop Mixup." Delve into Katherine M. Collins and colleagues' research on aligning model representations with human perceptions to improve robustness and generalization in machine learning. Discover their innovative approach using synthetic data from mixup, a powerful regularizer. Learn about the HILL MixE Suite, a series of elicitation interfaces designed to gather human perceptual judgments and uncertainties on mixup examples. Examine the findings that reveal discrepancies between human perceptions and traditional synthetic point labels. Gain insights into the potential applications of these discoveries for enhancing the reliability of downstream models, particularly when incorporating human uncertainty. Access the comprehensive slides and published paper for a deeper understanding of this cutting-edge research in artificial intelligence and human-computer interaction.