Robustness of Neural Networks to Parameter Quantization

Abhishek Murthy, Himel Das, Md Ariful Islam

Quantization, a commonly used technique to reduce the memory footprint of a neural network for edge computing, entails reducing the precision of the floating-point representation u…