Deep Learning Training on the Edge with Low-Precision Posits
This work addresses efficient DNN training for edge computing, but it is incremental as it extends posit usage from inference to training.
The paper tackles the problem of deep neural network training using low-precision posits instead of floating point, showing that 16-bit posits outperform 16-bit floating point in end-to-end training on MNIST and Fashion MNIST datasets.
Recently, the posit numerical format has shown promise for DNN data representation and compute with ultra-low precision ([5..8]-bit). However, majority of studies focus only on DNN inference. In this work, we propose DNN training using posits and compare with the floating point training. We evaluate on both MNIST and Fashion MNIST corpuses, where 16-bit posits outperform 16-bit floating point for end-to-end DNN training.