# Deep Universal Reinforcement Learning

- **Registry ID:** `catagi-deep-url`
- **Book:** Categories for AGI
- **Documentation status:** `curation-required`
- **Experimental status:** thirty-run controlled comparison
- **Canonical documentation URL:** https://categorical-ai.sridharmahadevan.com/experiments/catagi-deep-url

## Purpose

GT and diagrammatic-backpropagation ablations for coalgebraic RL losses.

## Book location

- Chapter 28: Deep URL with Geometric Transformers

## Documented studies

- Thirty-run base learner comparison
- Geometric Transformer structural-loss ablation
- Diagrammatic-backpropagation residual ablation
- Return, residual, and graph-smoothness comparison

## Associated code packages

- `deep-url-archive` — GT RL Coalgebra archive; **local-not-public**; license decision pending. Thirty-run Deep URL, GT, and diagrammatic-backpropagation experiment artifacts.
- `catagi-lean` — [Categories for AGI Lean companion](https://github.com/sridharmahadevan/catagi/tree/8ce79cdc949604873aca279d6cfcafa992293eb8); **public**; MIT. Machine-checked structural companion for selected categorical statements in Categories for AGI.

## Start here

These concrete files are selected from the complete resolved source surface. They orient the reader; they are not a claim that every family is independently reproducible.

- **runner · staged-not-public** — `deep-url-archive:run_synthetic_mdp_ablation.py`. Locate this file in the curated companion packet; no public download is currently offered.
- **runner · staged-not-public** — `deep-url-archive:run_synthetic_mdp_loop.py`. Locate this file in the curated companion packet; no public download is currently offered.
- **runner · staged-not-public** — `deep-url-archive:report_synthetic_mdp_ablation.py`. Locate this file in the curated companion packet; no public download is currently offered.
- **runner · staged-not-public** — `deep-url-archive:report_synthetic_mdp_metrics.py`. Locate this file in the curated companion packet; no public download is currently offered.

## Entry points

- `GT_RL_Coalgebra experiment scripts`

## Result artifacts

- `thirty-run return and residual summaries`

## Resolved code surfaces (5)

- **runner** — `deep-url-archive:report_synthetic_mdp_ablation.py`
- **runner** — `deep-url-archive:report_synthetic_mdp_metrics.py`
- **runner** — `deep-url-archive:run_synthetic_mdp_ablation.py`
- **runner** — `deep-url-archive:run_synthetic_mdp_loop.py`
- **runner** — `deep-url-archive:synthetic_mdp_bridge.py`

## Frozen run records (0)

- No frozen run-level record is registered. Consult the result-artifact list and evidence boundary above for the exact surviving evidence; absence of a run record is not evidence that no experiment was run.

## Evidence boundary

Lower structural residual or graph smoothness is not proof that a coalgebra morphism was learned.

A code or artifact link establishes traceability. It does not by itself establish independent reproduction, statistical adequacy, correctness, or support for a claim beyond this boundary.