Neural Analogical Matching

Analogy is core to human cognition. It allows us solve problems based on prior experience, it governs the way we conceptualize new information, and even influences our visual perception. The importance of analogy humans has made an active area research in broader field artificial intelligence, resulting data-efficient models that learn reason human-like ways. While cognitive perspectives deep learning have generally been studied independently one another, integration two lines a promising step towards more robust efficient techniques. As part growing body such integration, introduce Analogical Matching Network: neural architecture learns produce analogies between structured, symbolic representations are largely consistent with principles Structure-Mapping Theory.