Gravar-mail: A network model that can learn reward timing using reinforced expression of synaptic plasticity