In this paper, a sparsity-promoting adaptive algorithm for distributed learning in diffusion networks is developed. The algorithm follows the set-theoretic estimation rationale, i.e., at each time instant and at each node, a closed convex set, namely a hyperslab, is constructed around the current measurement point. This defines the region in which the solution lies. The algorithm seeks a solution in the intersection of these hyperslabs by a sequence of projections. Sparsity is encouraged in two complimentary ways: a) by employing extra projections onto a weighted ℓ1 ball, that complies with our desire to constrain the respective weighted ℓ1 norm and b) by adopting variable metric projections onto the hyperslabs, which implicitly quantify data mismatch. A combine-adapt cooperation strategy is adopted. Under some mild assumptions, the scheme enjoys a number of elegant convergence properties. Finally, numerical examples verify the validity of the proposed scheme, compared to other algorithms, which have been developed in the context of sparse adaptive learning.