Reinforcement learning for rule selection in end to end differentiable proving
Neural Theorem Provers (NTPs) are neuro-symbolic models that combine deep learning with a system of logic. They can learn representations for data, induce rules, are naturally interpretable, come with built-in explanations for conclusions, and demonstrate the capacity for systematic generalization....
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis |
Language: | English |
Published: |
2021
|
Subjects: |