Structural Generative Transmission (STB): A Bio-Inspired Framework for Transmitting Generative Instructions Instead of Raw Data
Year: 2026
Type: preprint
Abstract: We propose Structural Generative Transmission (STB), a bio-inspired framework for data compression and communication that transmits generative instructions instead of raw data. Inspired by molecular biology — DNA as a generative program, the ribosome as a fixed-primitive interpreter, the genetic code as a domain lookup table — STB separates a source into structural, substitutable, and irreducible residue components. The structural and substitutable parts are encoded as a program in a domain-specific metalanguage; the residue is classically compressed. A receiver equipped with a shared domain library executes the program to reconstruct the data locally. STB does not claim to circumvent Shannon's limits. We formalize the architecture using rate-distortion theory and Kolmogorov complexity, provide three formally proved propositions, hypothetical compression estimates with conditional ranges, and two concrete examples. A reference implementation (STB-Core v0.2, Rust) is available at https://doi.org/10.5281/zenodo.20683998. Empirical validation on real data is left as future work. Version 1.2.
Author: Matos Júnior, Daniel da Silva
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7163352521
Cites:
Cited by:
Related to: