Learning Program Synthesis for Integer Sequences from Scratch

We present a self-learning approach for synthesizing programs from integer sequences. Our method relies on a tree search guided by a learned policy. Our system is tested on the On-Line Encyclopedia of Integer Sequences. There, it discovers, on its own, solutions for 27987 sequences starting from basic operators and without human-written training examples.

Authors

• Thibault Gauthier Czech Technical University in Prague
• Josef Urban Czech Technical University in Prague

Keywords

• ML: Reinforcement Learning Algorithms
• ML: Transparent, Interpretable, Explainable ML
• ML: Unsupervised & Self-Supervised Learning
• SNLP: Language Models