# Gaussian Perceptual Field
**Entity class:** Proposed representational model
**Domain:** Computational Neuroscience / Neural Rendering / Perception
**Doc Type:** Canonical Concept Node
**Maturity:** Analytic synthesis
**Primary source:** [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]]
**Collection:** [[collections/Neurotech|Neurotech]]
---
## Definition
A **Gaussian perceptual field** is the proposed machine-side state representation inside the [[wiki/Closed-Loop Gaussian Sensorium Engine|Closed-Loop Gaussian Sensorium Engine]]. It represents a subject's current and forecasted perceptual state as localized, uncertain, temporally persistent primitives rather than as a finished pixel image.
## Two-layer model
1. **Retinotopic field:** Gaussian-like kernels carry visual-field or cortical position, anisotropic extent, luminance or color, salience, persistence, and confidence. This layer is designed to align receptive-field maps, phosphene maps, and stimulation effects.
2. **Ventral-temporal axis field:** distributed object-feature axes carry higher-order content shared between perception and imagery. The 2026 Wadia et al. study is an empirical anchor for shared perception–imagery coding; it does not by itself establish a Gaussian implementation in cortex.
The field is therefore a computational bridge between biological measurements and machine optimization. [[wiki/3D Gaussian Splatting|3D Gaussian Splatting]] supplies a useful renderer and differentiable primitive grammar, while the cortical analogy remains an article-level synthesis rather than a claim of literal identity.
## Relationships
- **Maintained by:** [[wiki/Closed-Loop Gaussian Sensorium Engine|Closed-Loop Gaussian Sensorium Engine]].
- **Rendered through:** [[wiki/3D Gaussian Splatting|3D Gaussian Splatting]].
- **Updated from:** [[wiki/Neural Signal Acquisition|Neural Signal Acquisition]] and [[wiki/Neural Decoding|Neural Decoding]].
- **Used by:** [[wiki/Perceptual Attractor Seeding|Perceptual Attractor Seeding]] to select sparse candidate interventions.
- **Interpreted through:** [[wiki/Predictive Processing|Predictive Processing]] and [[wiki/Active Inference|Active Inference]].
- **Governed by:** [[wiki/Perceptual Sovereignty|Perceptual Sovereignty]].
## Evidence boundary
Receptive-field models, phosphene simulators, neural decoding, and distributed object-axis codes are documented components. The combined two-layer field is a proposed integration. It must remain distinguishable from a measured neural object, a complete account of perception, or evidence that subjective experience has been reconstructed.
## Sources and provenance
- [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]] — source of the two-layer field and its role in the Engine.
- [Wadia et al., “A shared code for perceiving and imagining objects in human ventral temporal cortex” (Science, 2026)](https://pubmed.ncbi.nlm.nih.gov/41955351/).
- [van der Grinten et al., differentiable phosphene simulation (eLife, 2024)](https://elifesciences.org/articles/85812).
- [Kerbl et al., 3D Gaussian Splatting (SIGGRAPH 2023)](https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/).