A neuro-dynamic architecture for one shot learning of objects that uses both bottom-up recognition and top-down prediction Conference Paper