Seer proposes a machine learning-based predictor, enabling runtime kernel selection for irregular workloads with significant performance improvements.
