Lilach Bien

Class imbalance is a well-known challenge in both classical machine learning and deep learning. While numerous techniques exist to address it, selecting the most effective approach for a given problem remains complex. In this roundtable, we will begin the discussion by exploring how to determine whether an algorithm’s underperformance is truly caused by class imbalance, and which model types are more affected by it. Then we will discuss key strategies, including data sampling and loss adjustments, comparing their effectiveness across different scenarios. Attendees will gain a clearer understanding of when and how to apply these methods to improve model performance in real-world applications.