Similarity learning based on extension logic

Based on extension logic, this paper presents a novel learning method-similarity learning method. Similarity learning is a kind of learning driven by domain knowledge. The goal is to solve incompatible problems. It starts from the key characteristics of the goal and condition of original problems. During similarity learning, the extensibility is analyzed first, and then similarity goals of the original goals and corresponding similarity condition of original condition are considered. Finally, similarity transformations based on the principles of similarity transformations are made and thus the feasible satisfactory similarity solutions for similarity problems constitute similarity solutions for the original problems. Similarity learning has also a tradeoff between exploration and exploitation. The search process of similarity objects and similarity transformations is both a kind of trial-and-error search and data mining process. It differentiates from reinforcement learning in that it is expanded based on similarity biases and not on probability biases.