Abstract
This paper investigates the safety properties of inductive logic programming (ILP), particularly as compared to deep learning systems.
We consider the following properties: ease of model specification; robustness to input change; control over inductive bias; verification of specifications; post-hoc model editing; transfer learning; and interpretability.
We find that ILP satisfies many of these properties in some domains.
Lastly, we propose a hybrid system using ILP as a preprocessor to generate specifications for other ML systems.
We consider the following properties: ease of model specification; robustness to input change; control over inductive bias; verification of specifications; post-hoc model editing; transfer learning; and interpretability.
We find that ILP satisfies many of these properties in some domains.
Lastly, we propose a hybrid system using ILP as a preprocessor to generate specifications for other ML systems.
| Original language | English |
|---|---|
| Publication status | Accepted/In press - 2021 |
| Event | AAAI Workshop on Artificial Intelligence Safety 2021 - Duration: 8 Feb 2021 → … |
Workshop
| Workshop | AAAI Workshop on Artificial Intelligence Safety 2021 |
|---|---|
| Abbreviated title | SafeAI 2021 |
| Period | 8/02/21 → … |
Research Groups and Themes
- Interactive Artificial Intelligence CDT