Cooperative Multi-Point is a technology for improved spectrum utilization where a group of neighboring base stations coordinate their transmissions. However, the performance gain is severely degraded due to out-of-group interference. We propose simple schemes for null space learning and tracking of the interference channel between a terminal and an interfering base station group, whose only required network information is the interference levels at the terminal. In the proposed schemes, the interfering base station group transmits a sequence of learning signals which inflict low interference on a terminal of a neighboring group. The schemes utilize the fact that each terminal normally measures its interference levels and feeds them back to its serving base stations. Then, the serving group distributes this information to the interfering group, from which it learns the null space of the interfering channel. The proposed algorithms do not affect the transmission protocol between the terminal and the serving base stations, and do not add any signaling to the control channel between them. Each terminal communicates only with its serving group and treats the learning signals of the interfering group as noise.