Learn2Mix: Training Neural Networks Using Adaptive Data Integration
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908606406852608 |
|---|---|
| author | Venkatasubramanian, Shyam Tarokh, Vahid |
| author_facet | Venkatasubramanian, Shyam Tarokh, Vahid |
| contents | Accelerating model convergence in resource-constrained environments is essential for fast and efficient neural network training. This work presents learn2mix, a new 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 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_16482 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Learn2Mix: Training Neural Networks Using Adaptive Data Integration Venkatasubramanian, Shyam Tarokh, Vahid Machine Learning Accelerating model convergence in resource-constrained environments is essential for fast and efficient neural network training. This work presents learn2mix, a new 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 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. |
| title | Learn2Mix: Training Neural Networks Using Adaptive Data Integration |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2412.16482 |