Quantization Methods and Hardware Architectures for Neural Network and Big-Data Workloads