Combinatorial Manifold learning Based on Wolf Pack Algorithm with Gaussian Parameters

Isometric mapping algorithm and stochastic neighbor embedding algorithm are both important nonlinear dimension reduction algorithms in manifold learning algorithm, which are used to reduce the dimension of data from the perspective of popular distance and probability respectively, both of these algorithms will eventually come down to a nonlinear optimization problem, but usually the optimization methods we use, gradient descent and Stochastic gradient descent, can easily lead to the trap of local MINIMA, therefore, in order to obtain better global convergence and faster convergence speed, we need a better algorithm to replace it, that is, the combined manifold wolves algorithm introduced in this paper.