Design of Uniform Scalar Quantizer for Discrete Input Signals

This paper considers the design and performance of the uniform quantizer with discrete input, for the Laplacian source. Differences in the design process and performance, compared to the uniform quantizer with continual analog input are highlighted. Uniform quantizers with discrete input signals are largely used in practice, for quantization of the previously discretized signals, with the aim of compression. Especially important is possible application in DNNs (deep neural networks) for compression of DNN parameters.