Locally Linear Embedding Based on Faure Sequence Initialized Killer Whale Algorithm

In the field of artificial intelligence (AI), and to be specific, in the field of manifold learning, local linear embedding (LLE) algorithm, which is a nonlinear data dimensionality reduction method, has become an indispensable and important tool. The strength of LLE algorithm is that after dimensionality reduction, it can make the points in high-dimensional space isomorphic with the points in low-dimensional space after projection, and the algorithm has low complexity and fast operation speed. However, LLE also has its own drawback, it depends on gradient descent (GD) method or stochastic gradient descent (SGD) method in solving the optimization of its objective function, resulting in falling into the trap of local minimum. In the past, there was little discussion about the global optimum solution issue for LLE, therefore, in this paper, a Faure sequence initialized killer whale algorithm (FSKWA) has been proposed to effectively overcome the drawback in finding the global optimum solution of the objective function for LLE. And via numerical experiment, the superiority of the method proposed has been verified.