Index t-SNE Based on Neiderreit Sequence Initialized White Shark Optimizer

t-distributed stochastic neighbor embedding (t-SNE) is an unsupervised manifold learning method in the field of machine learning (ML). It mainly achieves noise removal and feature extraction through nonlinear data dimensionality reduction. However, t-SNE is a type of non parametric algorithm, the two initialization steps in its algorithm process may result in two different embeddings, leading to a decrease in computational efficiency. Index t-SNE method can solve the problem of t-SNE to some extent, but its loss function still relies on gradient line search methods to solve, such as gradient descent (GD) method and stochastic gradient descent (SGD) method which are prone to fall into the trap of local optima and cannot achieve global optimization. To overcome this drawback, in this paper, a derivative-free optimization method, which is called Neiderreit sequence initialized white shark optimizer (NSWSO) has been proposed. Via numerical experiment, its effectiveness and superiority comparing with gradient dependent methods are verified.