A “random-walk” simulation model of multiple-pattern learning in a radial-arm maze