Model-Code Separation Architecture for Data Compression Using Sum-Product Networks

Abstract

A decoding methodology that receives data compressed by a universal encoder and a data model based on a sum-product network (SPN) representing statistical structure inherent to source data. The approach enables decompression using this data model, implementing a framework where encoding remains independent of the source model while decoding adapts to model-specific information.

Type
Publication
US Patent App (US20260017501A1)