Self-Supervised Inductive Logic Programming

Researchers developed a new self-supervised Inductive Logic Programming approach called Poker that can learn recursive logic programs without requiring expert-crafted negative examples or problem-specific background theories. The system automatically generates and labels new training examples during learning, showing improved performance over existing methods when negative examples are unavailable.

• →New MIL algorithm called Poker learns from positive labeled examples and automatically generates negative examples during training.
• →The approach eliminates the need for expert-crafted background theories and negative examples that traditional ILP methods require.
• →Poker outperforms state-of-the-art Louise system in grammar learning tasks when negative examples are unavailable.
• →Introduces Second Order Definite Normal Form (SONF) for principled selection of background theories.
• →Performance improves with increasing numbers of automatically generated examples, addressing over-generalization issues.