English

Stochastic Markov Gradient Descent and Training Low-Bit Neural Networks

Machine Learning 2020-12-23 v2 Optimization and Control Machine Learning

Abstract

The massive size of modern neural networks has motivated substantial recent interest in neural network quantization. We introduce Stochastic Markov Gradient Descent (SMGD), a discrete optimization method applicable to training quantized neural networks. The SMGD algorithm is designed for settings where memory is highly constrained during training. We provide theoretical guarantees of algorithm performance as well as encouraging numerical results.

Keywords

Cite

@article{arxiv.2008.11117,
  title  = {Stochastic Markov Gradient Descent and Training Low-Bit Neural Networks},
  author = {Jonathan Ashbrock and Alexander M. Powell},
  journal= {arXiv preprint arXiv:2008.11117},
  year   = {2020}
}

Comments

19 pages, 2 figures

R2 v1 2026-06-23T18:05:44.369Z