Abstract: To kernelize an algorithm, there are two key points.That is, whether the algorithm can be kernelized and how to kernelize the algorithm if it can be kernelized.Generally it needs many mathematic skills to determine whether an algorithm can be kernelized.This paper provides a new determining condition for kernelization of algorithms.It has the advantage that it is not limited by the specific form of the algorithm.Traditional kernelizing methods are obtained through mapping the input data to the feature space and then running the algorithm in the feature space.This paper introduces another kernelizing method.The kernelizing method is equivalent to traditional kernelizing method through analysis and experiment.