Learning boolean halfspaces with small weights from membership queries

In this paper we close this gap and give an adaptive proper learning algorithm with two rounds, and asking nO(t) membership queries. We also give a non-adaptive proper learning algorithm that asks nO(t3) membership queries.