Neural Network Training with Approximate Logarithmic Computations
This addresses the problem of enabling online and real-time training on resource-constrained edge devices, though it appears incremental as it builds on existing log-domain approximations.
The paper tackles the high computational complexity of training deep neural networks on edge devices by proposing an end-to-end training and inference scheme using approximate logarithmic computations to eliminate multiplications. Their 16-bit log-based training achieves classification accuracy within approximately 1% of floating-point baselines on several common datasets.
The high computational complexity associated with training deep neural networks limits online and real-time training on edge devices. This paper proposed an end-to-end training and inference scheme that eliminates multiplications by approximate operations in the log-domain which has the potential to significantly reduce implementation complexity. We implement the entire training procedure in the log-domain, with fixed-point data representations. This training procedure is inspired by hardware-friendly approximations of log-domain addition which are based on look-up tables and bit-shifts. We show that our 16-bit log-based training can achieve classification accuracy within approximately 1% of the equivalent floating-point baselines for a number of commonly used datasets.