Random Reinforced Learning