Swarm-Based Gradient Descent Method for Efficient Global Optimization of Non-Convex Functions

The Swarm-Based Gradient Descent (SBGD) method is an effective global optimization algorithm for non-convex functions. It utilizes a swarm of communicating agents, where the agents dynamically adjust their step sizes and masses based on their relative positions and heights, enabling a simultaneous approach towards local minima while exploring for better global minima.