Quantization Modes for Neural Network Inference: ASIC Implementation Trade-offs | IEEE Conference Publication | IEEE Xplore