On the learning behavior of adaptive networks — Part II: Performance analysis