Conditional Random Fields Based on Neiderreit Sequence Initialized Eagle Perching Optimizer

Abstract

Conditional random field (CRF) is a spatial state model proposed by the probability graph school of thought, and it belongs to the undirected probability inference model in the area of probability graph model. CRF combines the advantages of several classical models in machine learning (ML) and plays an important role in solving the model parameters of ML algorithms which describe unconstrained optimization problems. However, CRF also have its own defects, for some functions, such as multi-modal non convex smooth objective functions, it is difficult for it to find the global optimal solution by using gradient dependent methods, and the solution process is easy to fall into local optimal solution. At present, many scholars have studied CRF related issues, but most of them have just tried to use CRF algorithm to solve specific application problems in various industries, there are few reports on the optimization of CRF algorithm itself, especially its solution process. Thus, in this paper, a derivative free optimization swarm intelligence method, namely Neiderreit sequence initialized eagle percolating optimizer (NSIEPO) has been proposed to replace the gradient dependent method in finding the global optimal solution. By numerical analysis, the effectiveness of the proposed algorithm has been verified.