Multi-Objective reward generalization: Improving performance of Deep Reinforcement Learning for applications in single-asset trading