Cognitive ad-hoc networks allow users to access an unlicensed/shared spectrum without the need for any coordination via a central controller and are being envisioned for futuristic ultra-dense wireless networks. The ad-hoc nature of networks require each user to learn and regularly update various network parameters such as channel quality and the number of users, and use learned information to improve the spectrum utilization and minimize collisions. For such a learning and coordination task, we propose a distributed algorithm based on a multi-player multi-armed bandit approach and novel signaling scheme. The proposed algorithm does not need prior knowledge of network parameters (users, channels) and its ability to detect as well as adapt to the changes in the network parameters thereby making it suitable for static as well as dynamic networks. The theoretical analysis and extensive simulation results validate the superiority of the proposed algorithm over existing state-of-the-art algorithms.