Randomized Matrix Sketching for Neural Network Training and Gradient Monitoring
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909818206289920 |
|---|---|
| author | Antil, Harbir Verma, Deepanshu |
| author_facet | Antil, Harbir Verma, Deepanshu |
| contents | Neural network training relies on gradient computation through backpropagation, yet memory requirements for storing layer activations present significant scalability challenges. We present the first adaptation of control-theoretic matrix sketching to neural network layer activations, enabling memory-efficient gradient reconstruction in backpropagation. This work builds on recent matrix sketching frameworks for dynamic optimization problems, where similar state trajectory storage challenges motivate sketching techniques. Our approach sketches layer activations using three complementary sketch matrices maintained through exponential moving averages (EMA) with adaptive rank adjustment, automatically balancing memory efficiency against approximation quality. Empirical evaluation on MNIST, CIFAR-10, and physics-informed neural networks demonstrates a controllable accuracy-memory tradeoff. We demonstrate a gradient monitoring application on MNIST showing how sketched activations enable real-time gradient norm tracking with minimal memory overhead. These results establish that sketched activation storage provides a viable path toward memory-efficient neural network training and analysis. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_00442 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Randomized Matrix Sketching for Neural Network Training and Gradient Monitoring Antil, Harbir Verma, Deepanshu Machine Learning 68T07, 65F55, 65K10, 68W20 Neural network training relies on gradient computation through backpropagation, yet memory requirements for storing layer activations present significant scalability challenges. We present the first adaptation of control-theoretic matrix sketching to neural network layer activations, enabling memory-efficient gradient reconstruction in backpropagation. This work builds on recent matrix sketching frameworks for dynamic optimization problems, where similar state trajectory storage challenges motivate sketching techniques. Our approach sketches layer activations using three complementary sketch matrices maintained through exponential moving averages (EMA) with adaptive rank adjustment, automatically balancing memory efficiency against approximation quality. Empirical evaluation on MNIST, CIFAR-10, and physics-informed neural networks demonstrates a controllable accuracy-memory tradeoff. We demonstrate a gradient monitoring application on MNIST showing how sketched activations enable real-time gradient norm tracking with minimal memory overhead. These results establish that sketched activation storage provides a viable path toward memory-efficient neural network training and analysis. |
| title | Randomized Matrix Sketching for Neural Network Training and Gradient Monitoring |
| topic | Machine Learning 68T07, 65F55, 65K10, 68W20 |
| url | https://arxiv.org/abs/2510.00442 |