t-SNE-PSO: Optimizing t-SNE using particle swarm optimization

t-distributed Stochastic Neighbor Embedding (t-SNE) is a manifold embedding technique that utilizes the Stochastic Gradient Descent (GD) algorithm to optimize objective functions to preserve pairwise distances between high-dimensional inputs in the lower-dimensional representation. Gradient-based methods are known by their local searches and may not explore the search space widely, making them susceptible to becoming trapped in local minima. To address this limitation, an adapted Particle Swarm Optimization (PSO) technique for t-SNE is proposed. The proposed t-SNE-PSO algorithm aims to overcome the limitations of GD by introducing a dynamic update of cognitive and social coefficients in PSO for optimizing t-SNE, enhancing its ability to find global optima and strike a balance between exploration and exploitation. Furthermore, the updated PSO contributes to developing a more efficient and effective dimensionality reduction technique, demonstrating superior qualities in clustering and visualization. The evaluation results on various benchmarks demonstrate the effectiveness of the proposed t-SNE-PSO algorithm.