A precise knowledge of the MIMO channel between the serving node and the user equipment (UE) is important for attaining good data rates in downlink transmissions (DL) in cellular systems. The interfering point-tomultipoint (I-P2MP) channel, consisting of multiple transmitters coexisting in the same area, where each transmitter is intended to serve multiple users, is a model that subsumes most of the scenarios that can be found in wireless cellular networks with a dense deployment of small cells (SCs). In conventional channel estimation procedures, resources allocated to each SC for training tend to be orthogonal, negatively impacting in the efficiency of the whole system. Reusing resources for training allows releasing resources for data transmission, but at the cost of degrading the channel estimation due to interference. We propose a decentralized algorithm for interference management that enforces the coordination among SCs in the design of the training sequences for de DL. Results in this work elucidate how to reuse resources for training and significantly improve the throughput of the system.