arXiv subjects

Jonathan Ashbrock

Publications and source records attributed to Jonathan Ashbrock.

2 recordsLinked to original sources

arXiv2020

Stochastic Markov Gradient Descent and Training Low-Bit Neural Networks

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.

cs.LG↗

arXiv2020

An Alternative Thresholding Rule for Compressed Sensing

Compressed Sensing algorithms often make use of the hard thresholding operator to pass from dense vectors to their best s-sparse approximations. However, the output of the hard thresholding operator does not depend on any information from a particular problem instance. We propose an alternative thresholding rule, Look Ahead Thresholding, that does. In this paper we offer both theoretical and experimental justification for the use of this new thresholding rule throughout compressed sensing.

math.NA↗