SYSTEMS AND METHODS FOR PERCEPTUAL QUALITY-DRIVEN ADAPTIVE QUANTIZATION IN NEURAL NETWORK DATA COMPRESSION WITH DYNAMIC FEEDBACK CONTROL

Abstract: a system for adaptive data compression uses content-aware analysis and dynamic feedback to optimize compression quality. an adaptive quantization subsystem analyzes content characteristics of input data and determines appropriate quantization parameters. a bit allocation engine distributes available bits across different portions of the input data based on the analyzed characteristics. a quality assessment subsystem monitors the compressed output and generates parameter adjustment signals based on measured quality metrics. a feedback control subsystem then modifies the quantization parameters in response to these signals. the modified parameters are used to create optimized compressed output data from the input dataset. this dynamic, content-aware approach enables improved compression quality while maintaining efficient data reduction.

Inventor(s): Zhu Li, Paras Maharjan, Brian Galvin

CPC Classification: G06N3/0455 (Auto-encoder networks; Encoder-decoder networks)

Search for rejections for patent application number 20250190765