We consider multi-armed bandit with distributed players, where each player independently samples one of N stochastic processes with unknown parameters and accrues reward in each slot without information exchange. Users choosing the same arm collide, and none or only one receives reward depending on the collision model. This problem can be formulated as a →