Neuro-Evolved Heuristics for Multi-Sequence Gapped Subsequence Identification

Researchers propose a neuro-evolutionary framework to automatically design neural network-based heuristics for the Variable Gapped Longest Common Subsequence Problem. The method optimizes weights using a genetic algorithm within an iterative multi-source beam search to improve robustness over hand-crafted designs in sequence alignment tasks.