Eidon – Neural State Geometry and Modelling Architecture: A Unified Architecture for Neural-State Representation, Machine Learning, and Multi-Framework Modelling

Large-scale analysis of neural-state-related data remains limited by heterogeneous repositories, acquisition systems, data formats, metadata conventions, preprocessing pipelines, feature definitions, representation schemes, and modelling outputs. These limitations restrict reproducibility and make it difficult to compare theoretical, computational, clinical, behavioural, cognitive, affective, contextual, social, and applied claims across datasets and research contexts. This paper specifies the architecture of Eidon: Neural State Geometry and Modelling Architecture, a declaratively configurable software system for neural-state representation, data harmonisation, feature extraction, multimodal alignment, machine learning, multi-framework modelling, evaluation, export, and provenance tracking. Eidon is centred on neural-state geometry and modelling, while allowing neural data to be represented, compared, and modelled in relation to associated physiological, behavioural, cognitive, clinical, contextual, social, digital-behavioural, intervention, longitudinal, demographic, and subject-level variables. Eidon is governed by declarative configuration files that specify data sources, enabled modules, preprocessing parameters, feature sets, representation options, modelling approaches, evaluation settings, export targets, and execution controls. Analysis runs are defined through configuration. The system is model-agnostic: any computational modelling approach may be included once it is integrated into Eidon’s configuration, execution, evaluation, export, and provenance workflow. Eidon produces reproducible analytical artefacts, including extracted measures, state representations, model outputs, predictions, forecasts, simulations, explanations, uncertainty estimates, causal and counterfactual analyses, decision-support artefacts, reports, visualisations, and provenance manifests. This paper defines the architecture, configuration semantics, module responsibilities, output schemas, and provenance guarantees required to support reproducible analysis, auditing, extension, and controlled use of the system.