DeepGraphLog for Layered Neurosymbolic AI
Adem Kikaj, Giuseppe Marra, Floris Geerts, Robin Manhaeve, Luc De Raedt · 2025
Read DeepGraphLog for Layered Neurosymbolic AI by Adem Kikaj, Giuseppe Marra, Floris Geerts, Robin Manhaeve, Luc De Raedt (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-07665.txt.
Description
Adem Kikaj and 4 co-authors present DeepGraphLog in 2025, a neurosymbolic framework that lets neural and symbolic layers alternate freely. Neurosymbolic AI joins neural learning with symbolic reasoning. It aims for both structure and interpretability. Systems such as DeepProbLog fix the flow, though. Neural processing always comes first, and symbolic reasoning second. That fixed order restricts complex dependencies, above all on graph-shaped data.
DeepGraphLog extends ProbLog with Graph Neural Predicates. Symbolic representations are treated as graphs, so graph neural networks can process them. Components can then stack in any order, across many layers. The team shows the framework on planning tasks. They also test it on knowledge graph completion with distant supervision, and on graph neural network expressivity. The authors report that the framework captures complex relational dependencies that fixed-flow systems miss.
Catalog record
| CREATOR | Adem Kikaj, Giuseppe Marra, Floris Geerts, Robin Manhaeve, Luc De Raedt |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-07665