Autonomous Observation Dynamics, Minimal All-Future Realizations, and Provenance Information Loss: From Terminating Quotient Refinement and Optimal Windows to Non-Resetting Physical Circuits
For a linear system over a commutative ring, we prove that an observation window admits an autonomous update if and only if its kernel equals the all-future invisible submodule. The criterion applies to every single-valued update map. For a fixed observer that reads all columns of a spatial, mass, and clock carrier, \(a+1\) consecutive frames recover every observable spatial coordinate and the independent mass and clock at copy scale \(a>1\). No update map on the full carrier can use fewer frames. A rational encoding and decoding protocol uses two moments. For finite systems with heterogeneous observation outputs, refining the old minimum quotient after adding a readout yields the same equivalence relation as recomputation. For streaming reception, counted moments make coarsening commute with updates and preserve the initial counted error across all finite continuations. One precision budget thus suffices for exact recovery of the evolving conditional distributions. For a fixed task, population, and weights, the excess coarsening risk is the mean squared distance between the fine and coarse conditional means. In an explicit parity-coupled family, retaining the same pair of channels yields a strict recovery gain at every successor, while replacement improves recovery of the new task at even depths \(d\ge2\) and degrades recovery of the previous task at odd depths \(d\ge1\). In both cases the worse channel has risk at least \((d+1)/(d+2)\). A dissipative RLC and gate-level model gives blind recovery of all 1024 independent drive configurations and a 128-bit reception program that completes by its deadline on every raw packet of the specified width. Along runs generated by the hardware, the receiver carries capacitor voltages across every finite sequence of frames without a dedicated reset-to-zero phase. For a decoder with access to one \(B\)-bit packet, the mean squared error for elapsed time over \(n\) frames is at least \(\tau^2[(n/2^B)^2-1]/12\), where \(\tau>0\) is a lower bound on frame duration.
Authors
- Jian Gao (ORCID: https://orcid.org/0009-0009-5002-1655)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24223414
- Primary Topic
- Cellular Automata and Applications
- Type
- preprint