Meta-learning via Search Combined with Parameter Optimization

摘要: Framework for Similarity-Based Methods (SBMs) allows to create many algorithms that differ in important aspects. Although no single learning algorithm may outperform other on all data an almost optimal be found within the SBM framework. To avoid tedious experimentation a meta-learning search procedure space of possible is used build new algorithms. Each generated by applying admissible extensions existing and most promising are retained extended further. Training performed using parameter optimization techniques. Preliminary tests this approach very encouraging.