| _version_ | 1866901169426661376 |
|---|---|
| author | Hussein, Ahmed Hadi |
| author_facet | Hussein, Ahmed Hadi |
| contents | <p><span>I introduce <strong>Topology-Rewiring Neural Cognitive Diffusion (TR-NCD)</strong>, a sequence modeling framework that <strong>redefines what a “weight” is</strong> and relocates long-term knowledge away from dense floating-point parameter matrices. In TR-NCD, I treat long-term knowledge as a <strong>structural graph</strong> (G=(V,E,m)) whose edges encode <strong>integer-valued masses/counts</strong> rather than real-valued multiplicative weights. I treat “weights” at runtime as <strong>addressed, ephemeral activation artifacts</strong>—identifiers that can be materialized into a temporary working set and then evicted immediately after use. This results in a clean separation between (i) <strong>structural parameters</strong> (graph topology + integer masses) as durable knowledge and (ii) <strong>context state</strong> (e.g., an SSM-like module) as temporal control logic for retrieval and ordering, rather than a repository of learned dense numeric weights.</span></p> <p><span>I formalize TR-NCD through a set of numbered definitions that specify: the symbol alphabet (including relation atoms), stable identifiers, ephemeral weight materialization, graph-based long-term storage, a thresholded diffusion operator, and training as local <strong>topology rewiring</strong> rather than backpropagated gradient updates on floating-point matrices. In later sections of the paper (continued in subsequent parts), I extend the model with <strong>Expandable Relation Keys (ERK)</strong> for variable-order relational contexts and an <strong>Ambiguity Decision Manager (ADM)</strong> that explicitly handles uncertainty via multi-hypothesis, deferred commitment, and an “ask/expose” mode instead of forced single-path output. I also provide complete algorithmic specifications for inference, training, on-demand creation, and maintenance/pruning, as well as a concrete evaluation plan and ablation matrix template.</span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18216805 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Topology-Rewiring Neural Cognitive Diffusion (TR-NCD) — v1.0 Hussein, Ahmed Hadi Artificial Intelligence Cognitive Modeling Diffusion Models Graph-based Learning Graph Neural Networks Machine Learning Neural Architecture Structural Learning Topology Rewiring TR-NCD <p><span>I introduce <strong>Topology-Rewiring Neural Cognitive Diffusion (TR-NCD)</strong>, a sequence modeling framework that <strong>redefines what a “weight” is</strong> and relocates long-term knowledge away from dense floating-point parameter matrices. In TR-NCD, I treat long-term knowledge as a <strong>structural graph</strong> (G=(V,E,m)) whose edges encode <strong>integer-valued masses/counts</strong> rather than real-valued multiplicative weights. I treat “weights” at runtime as <strong>addressed, ephemeral activation artifacts</strong>—identifiers that can be materialized into a temporary working set and then evicted immediately after use. This results in a clean separation between (i) <strong>structural parameters</strong> (graph topology + integer masses) as durable knowledge and (ii) <strong>context state</strong> (e.g., an SSM-like module) as temporal control logic for retrieval and ordering, rather than a repository of learned dense numeric weights.</span></p> <p><span>I formalize TR-NCD through a set of numbered definitions that specify: the symbol alphabet (including relation atoms), stable identifiers, ephemeral weight materialization, graph-based long-term storage, a thresholded diffusion operator, and training as local <strong>topology rewiring</strong> rather than backpropagated gradient updates on floating-point matrices. In later sections of the paper (continued in subsequent parts), I extend the model with <strong>Expandable Relation Keys (ERK)</strong> for variable-order relational contexts and an <strong>Ambiguity Decision Manager (ADM)</strong> that explicitly handles uncertainty via multi-hypothesis, deferred commitment, and an “ask/expose” mode instead of forced single-path output. I also provide complete algorithmic specifications for inference, training, on-demand creation, and maintenance/pruning, as well as a concrete evaluation plan and ablation matrix template.</span></p> |
| title | Topology-Rewiring Neural Cognitive Diffusion (TR-NCD) — v1.0 |
| topic | Artificial Intelligence Cognitive Modeling Diffusion Models Graph-based Learning Graph Neural Networks Machine Learning Neural Architecture Structural Learning Topology Rewiring TR-NCD |
| url | https://doi.org/10.5281/zenodo.18216805 |