Online non-convex learning for river pollution source identification

In this article, novel gradient-based online learning algorithms are developed to investigate an important environmental application: real-time river pollution source identification, which aims at estimating the released mass, location, and time of a based on downstream sensor data monitoring concentration. The is assumed be instantaneously once. problem can formulated as non-convex loss minimization in statistical learning, our have vectorized adaptive step sizes ensure high estimation accuracy three dimensions different magnitudes. order keep algorithm from sticking saddle points loss, “escaping points” module multi-start setting derived further improve by searching for global minimizer functions. This shown theoretically experimentally O(N) local regret probability cumulative bound under particular error condition A real-life identification example shows superior performance compared with existing methods terms accuracy. Managerial insights decision maker use also provided.