PhD student, Duke University
1 paper at NeurIPS 2025
This work introduces learn2mix, a new training strategy that adaptively adjusts class proportions in batches to accelerate neural network convergence in resource-constrained environments.