Effective Quantization Methods for Recurrent Neural Networks
arXiv: Learning, Volume abs/1611.10176, 2016.
Reducing bit-widths of weights, activations, and gradients of a Neural Network can shrink its storage size and memory usage, and also allow for faster training and inference by exploiting bitwise operations. However, previous attempts for quantization of RNNs show considerable performance degradation when using low bit-width weights and a...More
Full Text (Upload PDF)
PPT (Upload PPT)