Alien Coding
This addresses the challenge of automating program synthesis for mathematical sequences, which is incremental as it builds on existing neural translation methods.
The authors tackled the problem of automatically synthesizing programs for OEIS sequences using a self-learning algorithm that starts from random programs and iteratively trains a neural machine translation model to propose new programs, resulting in the discovery of programs for over 78,000 sequences.
We introduce a self-learning algorithm for synthesizing programs for OEIS sequences. The algorithm starts from scratch initially generating programs at random. Then it runs many iterations of a self-learning loop that interleaves (i) training neural machine translation to learn the correspondence between sequences and the programs discovered so far, and (ii) proposing many new programs for each OEIS sequence by the trained neural machine translator. The algorithm discovers on its own programs for more than 78000 OEIS sequences, sometimes developing unusual programming methods. We analyze its behavior and the invented programs in several experiments.