Amortized Q Learning Patent - GDM Holding LLC - EP3867821A1

EPO published patent application EP3867821A1 for Amortized Q Learning, a machine learning technique for controlling agents in reinforcement learning environments. The patent application was filed by GDM Holding LLC with inventors including Tom Van de Wiele, Volodymyr Mnih, Andriy Mnih, and David Constantine Patrick Warde-Farley. The application covers multiple G06N subclasses including neural network learning methods (G06N 3/006, 3/045, 3/047, 3/084, 3/092).

Technology companies developing AI systems, reinforcement learning applications, or neural network-based control systems should review the patent claims for potential licensing requirements or freedom-to-operate considerations. The patent's designation across 33 European states means it could affect operations in Germany, France, United Kingdom, Italy, Spain, Sweden, and other EPO member states.