Utility Maximization in a Market with Competitive Heterogeneous Agents: backward propagation of chaos and learning