Learn2Mix: Training Neural Networks Using Adaptive Data Integration

Accelerating model convergence within resource-constrained environments is critical to ensure fast and efficient neural network training. This work presents learn2mix, a novel training strategy that adaptively adjusts class proportions within batches, focusing on classes with higher error rates. Unlike classical training methods that use static class proportions, learn2mix continually adapts class proportions during training, leading to faster convergence. Empirical evaluations conducted on benchmark datasets show that neural networks trained with learn2mix converge faster than those trained with existing approaches, achieving improved results for classification, regression, and reconstruction tasks under limited training resources and with imbalanced classes. Our empirical findings are supported by theoretical analysis.

@article{venkatasubramanian2025_2412.16482,
  title={ Learn2Mix: Training Neural Networks Using Adaptive Data Integration },
  author={ Shyam Venkatasubramanian and Vahid Tarokh },
  journal={arXiv preprint arXiv:2412.16482},
  year={ 2025 }
}